Lattice Based Cryptography for Beginners

– A supplementary note to the following

1. Peikert’s Bonn Lecture Slides

2. Lyubashevsky, Peikert and Regev:
A toolkit for Ring-LWE
3. Steinfeld’s Lecture Slides on multilinear maps
with Cryptanalysis of GGH map due to Hu and Jia

Dong Pyo Chi1,2 , Jeong Woon Choi3 , Jeong San Kim4 and Taewan Kim5
1
Division of General Studies, UNIST
[email protected]
2
Department of Mathematics, Seoul National University
[email protected]
3
Fusion Technology R&D Center, SK Telecom
jw [email protected]
4
Department of Applied Mathematics, Kyung Hee University
[email protected]
5
Institute of Mathematical Sciences, Ewha Womans University
[email protected]
Abstract

The purpose of this lecture note is to introduce lattice based cryptography, which is
thought to be a cryptosystem of post-quantum age. We have tried to give as many details
possible specially for novice on the subject. Something may be trivial to an expert but
not to a novice.
Many fundamental problems about lattice are thought to be hard even against quan-
tum computer, compared to factorization problem which can be solved easily with quan-
tum computer, via the celebrated Shor factorization quantum algorithm. The first part of
our presentation is based on slides of Christ Peikert 2013 Bonn lecture (crypt@b-it2013).
We, more or less, give somewhat detailed explanation of Professor Peikert’s lecture slides.
We unfortunately could not attend his Bonn class. We are afraid that there are many
mistakes in this note; if any, they are due to our misunderstanding of the material. Part
II of our lecture note is on ring LWE, based on the paper “A tool-kit for Ring-LWE
Cryptography” by Lyubashevsky, Peikert and Regev. Part III is about multilinear maps
together with cryptanalysis of GGH map due to Hu and Jia. Our presentation follows
professor Steinfeld’s lecture slides on GGHLite, and the paper by Yupu Hu and Huiwen
Jia. When you read this lecture note, the corresponding original paper should be ac-
companied. We thank professor Jung Hee Cheon for introducing the subject and asking
Dong Pyo Chi to give a lecture on the subject at the department of mathematics in
Seoul National University. We also thank Hyeongkwan Kim for many helps, especially
many corrections and improvements of the manuscript during the 2015 Summer session
at UNIST. We also thank the students who took the classes at SNU and UNIST. The
lecture was given by a novice for novice, so many mistakes are unavoidable. If the reader
lets us know any errors, we will very much appreciate it.

i
ii
Contents

Abstract

I Lattice Based Cryptography vii

1 Mathematical and Computational Background 1
1.1 Mathematical Background . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Two simple bounds on the minimum distance . . . . . . . . . . . 4
1.2 Computational Background . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.1 Hard problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

2 Short Integer Solution and Learning With Errors 9

2.1 Hard problems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 Short Integer Solution . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.2 Learning With Errors . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Cryptosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1 Public-Key Cryptosystem using LWE . . . . . . . . . . . . . . . . 12
2.2.2 Dual cryptosystem . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 More efficient Cryptosystem . . . . . . . . . . . . . . . . . . . . . 13

3 Discrete Gaussians and Applications 15

3.1 Discrete Gaussians and sampling . . . . . . . . . . . . . . . . . . . . . . 15
3.1.1 Discrete Gaussians . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.2 Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2.1 Identity Based Encryption . . . . . . . . . . . . . . . . . . . . . . 20

4 Constructing Trapdoors and More Applications 21

4.1 Strong trapdoor generation and inversion algorithms . . . . . . . . . . . 21
4.1.1 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
4.2 Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

II Introduction to Ring-LWE 27
5 Preliminaries for Ring-LWE cryptography 29
5.1 Notations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
5.2 Gaussians and Subgaussian Random Variables . . . . . . . . . . . . . . . 30

iii
 5.3 Lattice Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
5.3.1 Decoding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5.4 Algebraic Number Theory Background . . . . . . . . . . . . . . . . . . . 34
5.4.1 A key fact from algebraic number theory . . . . . . . . . . . . . . 35
5.4.2 Canonical Embedding and Geometry . . . . . . . . . . . . . . . . 35
5.4.3 The Ring of Integers and Its Ideals . . . . . . . . . . . . . . . . . 36
5.4.4 Duality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
5.4.5 Prime Splitting and Chinese Remainder Theorem . . . . . . . . . 40
5.5 Ring-LWE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

6 Discrete Fourier Transform & Chinese Remainder Transform 45

7 Powerful basis 47
7.1 Powerful basis p⃗ of K = Q(ζm ) and R = Z[ζm ] . . . . . . . . . . . . . . . 47
7.2 Gram-Schmidt orthogonalization of CRTm . . . . . . . . . . . . . . . . . 49

8 Chinese Remainder Basis and Fast Ring Operation 51

9 Decoding Basis of R∨ 53
9.1 Relation to the Powerful Basis . . . . . . . . . . . . . . . . . . . . . . . . 53
9.2 Decoding R∨ and its Powers . . . . . . . . . . . . . . . . . . . . . . . . . 54
9.2.1 Implementation of Decoding Operation . . . . . . . . . . . . . . . 55
9.3 Gaussian sampling in the Decoding Basis . . . . . . . . . . . . . . . . . . 56

10 Regularity 59

11 Cryptosystems 65
11.1 Dual-Style Cryptosystem [GPV08] . . . . . . . . . . . . . . . . . . . . . . 65
11.2 Compact Public-key Cryptosystem . . . . . . . . . . . . . . . . . . . . . 66
11.3 Homomorphic Cryptosystem . . . . . . . . . . . . . . . . . . . . . . . . . 67
11.3.1 Modulus Reduction and Key Switching . . . . . . . . . . . . . . . 68

III Multilinear map 75

12 Multilinear maps 77
12.1 Why multilinear map? . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
12.2 Grag-Gentry-Halevi (GGH) Graded Encoding Scheme . . . . . . . . . . . 78

13 GGHLite scheme for k-graded encoding 83

14 Cryptanalysis of GGH map 87

14.1 Schematic description of the cryptanalysis . . . . . . . . . . . . . . . . . 87
14.2 Generating an equivalent secret . . . . . . . . . . . . . . . . . . . . . . . 87
14.3 Modified Encoding/Decoding . . . . . . . . . . . . . . . . . . . . . . . . 88
14.4 Witness encryption based on 3–exact cover . . . . . . . . . . . . . . . . . 88
14.5 Breaking WE based on the hardness of 3–exact cover problem . . . . . . 89
14.6 Computing the Hermite
⟨ Normal Form⟩ of⟨ ⟨g⟩ by computing⟩ the Hermite
K−2 (1)
Normal Forms of h(1 + ag) b and h(1 + ag)K−2 b(1) g . . . . . . . 93

iv
 A Hermite Normal Form of Ideal Lattices (following Ding and
Lindner, Smart and Vercauteren) 95

B Notes on Cyclotomic Fields with Examples (by H. Kim) 97

B.1 Cyclotomic Number Fields & Ring of Integers . . . . . . . . . . . . . . . 97
B.2 The Space H and the Canonical Embedding . . . . . . . . . . . . . . . . 98
B.2.1 The Space H . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
B.2.2 The Canonical Embedding . . . . . . . . . . . . . . . . . . . . . . 99
B.3 Discriminant . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
B.4 Ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
B.4.1 Fractional ideals . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

v
vi
 Part I

Lattice Based Cryptography

vii
Chapter 1

Mathematical and Computational

Background

1.1 Mathematical Background

In Part I, we use the notations in [P13].

1.1.1 Definitions
Lattice
A lattice L of Rn is by definition a discrete subgroup of Rn . In this note we only deal
with full-rank lattice, i.e., L spans Rn with real coefficients. Moreover, we consider only
integer lattices, i.e., L ⊆ Zn .
√
Remark 1.1.1. Z + 2Z is not a lattice. Note that when α is irrational, nα mod 1 is
uniformly dense in S 1 = [0, 1]/0 ∼ 1 (Weyl theorem).

Bases
A basis of L is an ordered set B = (b1 , b2 , . . . , bn ) such that
{ n }
∑
L = L(B) = B · Zn = ci bi : ci ∈ Z . (1.1)
i=1

Note that by convention, bi are column vectors and B · k = k1 b1 + · · · + kn bn , where k

is a column vector.
Fundamental parallelepiped of basis B is
[ )n
1 1
P (B) = B · − , (1.2)
2 2
{ n }
∑ 1 1
= αi bi : − ≤ αi < . (1.3)
i=1
2 2

Note that P (B) depends not only on lattice but also on the choice of basis B. A “good”
basis of L gives rather a square-like parallelepiped, while a ‘bad’ basis gives a very thin
parallelepiped. It is trivial to see the following lemma.

1
Lemma 1.1.2. ∪
Rn = (v + P (B)), (1.4)
v∈L

that is, parallel translation by lattice vectors of parallelepiped covers Rn without overlap.

Proof. For any p ∈ Rn ,

∑
p = xi bi (1.5)
i
∑ ∑
= ⌈xi ⌋bi + (xi − ⌈xi ⌋)bi , (1.6)
i i

where ⌈a⌋ means rounding off. For example, ⌈2.7⌋ = 3, ⌈2.5⌋ = 3, and ⌈2.1⌋ = 2.
Therefore,
1 1
− ≤ a − ⌈a⌋ < . (1.7)
2 2
∑ ∑ ∪
Hence, ⌈xi ⌋bi ∈ L and (xi − ⌈xi ⌋)bi ∈ P (B). This shows that Rn = v∈L (v +
i i
P (B)).
If (v1 + P (B)) ∩ (v2 + P (B)) ̸= ∅ for some v1 ̸= v2 ∈ L, then v1 + α = v2 + β for
some α, β ∈ P (B), so v1 − v2 = β − α. Since v1 − v2 is a Z-linear combination of bi
while β − α is a (−1, 1)-linear combination of bi , so v1 − v2 = 0 = β − α.

BU is also basis for any U ∈ GL(n : Z), i.e., U is an n × n integer matrix with
determinant ±1. Note that, for example,
( )
1 1023
∈ GL(2 : Z). (1.8)
0 1

Coset and Determinant

It is much better to think a coset element of Zn /L concretely (note that we assumed
L ⊆ Z n ), as a subset v + L, i.e. a shift of the lattice L, where v ∈ Zn represents a coset
of Zn /L.

Lemma ∪ 1.1.3. Each coset of L has a unique representative in a parallelepiped P (B),

because (v + P (B)) covers Rn without overlap.
v∈L
∪
Proof. Let v ∈ Zn be a representative of a coset v + L. Since (w + P (B)) covers
w∈L
Rn without any overlap, there exists a unique w ∈ L such that v ∈ (w + P (B)). Then
v − w ∈ P (B), and v represents the same coset, i.e.,

v + L = (v − w) + L, (1.9)

so v − w is a representative of the coset v + L in P (B). Moreover, such a representative

is unique, since if v1 , v2 ∈ P (B) and

v1 + L = v2 + L, (1.10)

2
where
∑ 1 1
v1 = c1j bj , − ≤ c1j < , (1.11)
2 2
∑ 1 1
v2 = c2j bj , − ≤ c2j < , (1.12)
2 2
then ∑
v1 − v2 = (c1j − c2j )bj ∈ L, (1.13)
i.e., c1j − c2j ∈ Z for all j. Note that if − 12 ≤ a < 12 and − 21 ≤ b < 1
2
, then then
−1 ≨ a − b ≨ 1. Hence, c1j − c2j = 0 for j = 1, 2, . . . , n.
By definition,
det(L) := |Zn /L| = | det B| = vol(P (B)) (1.14)
for any basis B of L.

Lemma 1.1.4. |Zn /L| = vol(P (B)).

Proof. Note the following:

• L + P (B) covers Rn without overlap.

• Z
[ 1+ 1 )·n covers R without overlap, where · means the half closed unit cube
n n

−2, 2 .

Thus,

L + P (B) = Rn (1.15)
= Zn + · (1.16)
∪
= (c + L + · ). (1.17)
c ∈ Zn/L

It follows that |Zn /L| | · | = |P (B)|, so |Zn /L| = vol(P (B)).

Successive Minima
Successive minima of linearly independent vectors are defined as follows:

• λ1 (L) := min0̸=v∈L ∥v∥ = minx̸=y∈L ∥x − y∥

• λi (L) := min {r : L contains i linearly independent vectors of length ≤ r} .

Then λ1 (L) ≤ λ2 (L) ≤ · · · ≤ λn (L). Let v1 , v2 , . . . , vn be corresponding lattice elements.

Note that {v1 , v2 , . . . , vn } need not be a basis of L.

Example 1.1.1. Let L ⊂ Zn be spanned by 2e1 , . . . , 2en , (1, 1, . . . , 1), where n > 4.
Then v = (v1 , . . . , vn ) ∈ L if and only if v1 = v2 = · · · = vn mod 2. Then

λ1 (L) = · · · = λn (L) = 2. (1.18)

But {2e1 , . . . , 2en } is not a basis of L. (1, 1, . . . , 1) or its variation should be an element
of any basis of L.

3
1.1.2 Two simple bounds on the minimum distance
Gram-Schmidt Orthogonalization and Lower Bounding λ1
e of a basis B of L is given by
The Gram-Schmidt orthogonalization B
B = QR (1.19)
 
f1 ∥
∥b ∗
 .. 
= Q .  (1.20)
0 fn ∥
∥b
 
1 ∗
e
= B  ..
.

, (1.21)
0 1
where  
f1 ∥
∥b 0
e = Q
B  ... 
,
0 fn ∥
∥b
and Q is an orthonormal basis reduced from B, e and R is a representation of B with
respect to this basis.
[ )
Lemma 1.1.5. P (B) e =B e · − 1 , 1 n is a fundamental domain of L. That is, L + P (B) e
2 2
covers R without overlap.
n

e = vol(P (B)), it suffices to show that there is no overlap. Assume

Proof. Since vol(P (B))
there is a overlap, i.e.,
Bx + Bα e = By + Bβ e (1.22)
[ )
for some x, y ∈ Zn and α
n
⃗ , β⃗ ∈ − 12 , 21 . Then B(x − y) = B( e β⃗ − α
⃗ ). Letting z = x − y,
 
1 ∗
e  .  e β⃗ − α
B ..  z = B( ⃗ ), (1.23)
0 1
so  
1 ∗
 .. 
 .  z = (β⃗ − α
⃗ ). (1.24)
0 1
Note that z is an integer vector and
−1 ≨ βi − αi ≨ 1. (1.25)
From the equality (1.24),
zn = βn − αn ∴ zn = 0 → αn = βn (1.26)
zn−1 + ∗zn = βn−1 − αn−1 (1.27)
zn−1 = βn−1 − αn−1 ∴ zn−1 = 0 → αn−1 = βn−1 (1.28)
···
∴ z1 = 0 (1.29)
i.e., x = y. (1.30)

4
 It is easy to see that λ1 (L) ≥ min ∥bei ∥ from Bc = Q(Rc) for c ∈ Zn .
i

Upper Bounding λ1 : Minkowski’s Theorem

Theorem 1.1.6 (Minkowski Theorem 1). Any convex centrally symmetric body S of
volume > 2n det(L) contains a nonzero lattice point.
Proof. Let S ′ = 12 S, so vol(S ′ ) > det(L). Then there exist x ̸= y ∈ S ′ such that
x − y ∈ L, since for some v1 ̸= v2 ∈ L,
∩
(v1 + S ′ ) (v2 + S ′ ) ̸= ϕ (1.31)
z = v1 + x = v2 + y, x, y ∈ S ′ (1.32)
x − y = v2 − v1 ̸= 0 ∈ L. (1.33)

Now 2x, −2y ∈ S by the definition of S ′ , so

1 1
x − y = (2x) + (−2y) ∈ S
2 2
by the convexity of S.
Corollary 1.1.7. √ 1
λ1 (L) ≤ n(det L) n . (1.34)
Proof. We give a proof of the corollary using the following two facts:
√ 1
• A ball of radius > n(det L) n is convex and centrally symmetric.
√ 1 1
• B(0, n(det L) n ) ⊃ a cube of side length 2(det L) n , since
√
dist ((1, . . . , 1), (0, . . . , 0)) = n.

It follows that √ 1
vol(B(0, n(det L) n )) > 2n det L.

Remark 1.1.8. We could obtain a more refined inequality if we use the exact formula
for vol(B(0, R)). Choose R such that vol(B(0, R)) = 2n det L. Then λ1 (L) ≤ R.
∏ 1 √ 1
Theorem 1.1.9 (Minkowski Theorem 2). ( ni=1 λi (L)) n ≤ n(det L) n .
Proof. We may assume ∥bi ∥ = λi (L) for i = 1, . . . , n, and consider a lattice generated by
b1 , . . . , bn , possibly a sublattice of L.
 ⟨ ⟩ 2 

 e 

∑n y, bi
T := y ∈ R : n   <1 . (1.35)

 ∥ ei ∥λi
b 

i=1

Claim: The ellipsoid T does not contain any nonzero lattice point.
Let 0 ̸= y ∈ L, and 1 ≤ k ≤ n maximal such that

λk+1 (L) ≩ ∥y∥ ≥ λk (L). (1.36)

5
We claim y ∈ span{b1 , . . . , bk } = span{b̃1 , . . . , b̃k }. If not, b1 , . . . , bk , y are k+1 linearly
independent and their norms are less than λk+1 , a contradiction. Hence,
⟨ ⟩ 2 ⟨ ⟩ 2
∑ n y, b̃i ∑ k y, b̃i
  =   (1.37)
i=1
∥ b̃i ∥λ i i=1
∥ b̃i ∥λ i
⟨ ⟩ 2
∑ k
1  y, b̃i
≥  (1.38)
i=1
λ 2
k ∥ b̃ i ∥
∥y∥2
= ≥ 1, (1.39)
λ2k
so y ∈
/ T , i.e., T does not contain any nonzero lattice vector. Hence,
( n ) ( n )( )n
∏ ∏ 2
2 det(L) ≥ vol(T ) =
n
λi vol(B(0 : 1)) ≥ λi √ , (1.40)
i=1 i=1
n
so ( ) n1
∏
n
√ 1
λi ≤ n(det L) n . (1.41)
i=1

1.2 Computational Background

1.2.1 Hard problems
Shortest Vector Problem (SVP)
• SV Pγ : Given a basis B of L, find nonzero v ∈ L such that ∥v∥ ≤ γλ1 (L).
There exists an exact algorithm for finding a nonzero minimum vector in time 2O(n) ,
polynomial time algorithm for gap 2n , but no quantum algorithm with exponential
boost.

• GapSV Pγ : Given a basis B of L and a real d, decide between λ1 (L) ≤ d and

λ1 (L) > γd.

Note that GapSV Pγ ≤ SV Pγ , i.e., GapSV Pγ reduces to SV Pγ . (SV Pγ → find v ̸= 0 ∈ L

such that λ1 (L) ≤ ∥v∥ ≤ γλ1 (L).) If ∥v∥ ≤ γd, then λ1 ≤ γd. Hence, λ1 < d (because
either λ1 ≤ d or λ1 > γd). If ∥v∥ > γd, then γd < ∥v∥ ≤ γλ1 , so d < λ1 , hence λ1 > γd.)
But the reverse direction is open.

6
LLL (Lenstra-Lenstra-Lovaz) algorithm
B = (b1 , . . . , bn ) is a δ − LLL reduced basis if

(i) For 1 ≤ j < i ≤ n, we have |µi,j | ≤ 12 .

(ii) For 1 ≤ i < n, we have

δ∥b̃i ∥2 ≤ ∥µi+1,i b̃i + b̃i+1 ∥2 = |µi+1,i |2 ∥b̃i ∥2 + ∥b̃i+1 ∥2 , (1.42)

where
⟨ ⟩
bi , b̃j
µi,j = , (1.43)
∥b̃j ∥2
∑i−1
b̃i = bi − µi,j b̃j . (1.44)
j=1

B has the following form with respect to the orthonormal basis

 
∥b̃1 ∥ µ2,1 ∥b̃1 ∥ µ3,1 ∥b̃1 ∥ ∗ ≤ 12 ∥b̃1 ∥
 ∥b˜2 ∥ µ3,2 ∥b̃1 ∥ ∗ ≤ 12 ∥b˜2 ∥ 
 
 .. .. 
 . .  (1.45)
 
 ˜ ∥
≤ 2 ∥bn−1
1 
∥b̃n ∥

In particular, if 1 ≥ δ > 14 , then

( )
1
∥b̃i+1 ∥ ≥ δ −
2
∥b̃i ∥2 . (1.46)
4

Hence,
∥b1 ∥ = ∥b̃1 ∥ ≤ 2(n−1)/2 min ∥b̃i ∥ ≤ 2(n−1)/2 λ1 (L).
(We choose δ = 43 .)

7
LLL-algorithm

• Input: Lattice basis b1 , . . . , bn ∈ Zn .

• Output: δ-LLL reduced basis of L.

• Start: compute the Gram-Schmidt Orthogonalization b̃1 , b̃2 , . . . , b̃n .

• Reduction Step:
for i = 2 to n do
for j = i − 1 to 1 do ⌈
⟨bi ,b̃j ⟩ ⌋
bi ← bi − cij bj , where cij = .
⟨b̃j ·b̃j ⟩

• Swap Step:
If ∃ i such that δ∥b̃i ∥2 > ∥µi+1,i b̃i + b̃i+1 ∥2
bi ↔ bi+1
goto start.

• Output: b1 , . . . , bn .

Shortest Independent Vectors Problem (SIV Pγ )

Given a basis B, find linearly independent vectors v1 , . . . , vn ∈ L such that ∥vi ∥ ≤
γλn (L).

Bounded-Distance Decoding (BDD)

Given a basis B, ⃗t ∈ Rn , and real d < λ1 /2 such that dist(⃗t, L) ≤ d, find the unique
v ∈ L closest to ⃗t. BDD is equivalent to finding e ∈ ⃗t + L such that ∥e∥ ≤ d.

Algorithms for BDD

1. Babai’s Round off algorithm for BDD
Using a good basis B,
∑ ∑
⃗t = αi bi → e := (αi − ⌈αi ⌋)bi . (1.47)

It works if Ball(d) ⊆ P (B). Hence, d ≤ min∥b⊥ ⊥

i ∥/2, where bi is the orthogonal
component of bi to the hyperplane span{b1 , · · · , b̂i , · · · , bn }.

2. Babai’s nearest plane algorithm for BDD

Output e = ⃗t mod B,e where e ∈ P (B).
e
e where B
It works if Ball(d) ⊆ P (B), e is the Gram-Schmidt Orthogonalization of B
as before,
i.e., d ≤ min ∥bei ∥/2. (1.48)
i

e is also a fundamental domain of the lattice L.

Note that P (B)

8
Chapter 2

Short Integer Solution and Learning

With Errors

2.1 Hard problems

2.1.1 Short Integer Solution
Short Integer Solution (SIS)
Znq := n-dimensional vectors modulo q. Given a⃗1 , . . . , a⃗m ∈ Znq , find nontrivial and small
z1 , . . . , zm ∈ Z such that
z1 a⃗1 + · · · + zm a⃗m = 0 (2.1)
in Znq , i.e.,
Az = 0 mod q, (2.2)
where A = (⃗a1 , . . . , ⃗am ). This is finding a short vector in the lattice
( )
⊥ A∈Zn×m
L(A) := ker Zm q
/ Znq
z7→Az

= {x ∈ Zm : Ax = u mod q}.

One-way Hash Function

Set m > n log q. Define fA : {0, 1}m → Znq as

fA (x) = Ax. (2.3)

Then fA covers Zqn almost uniformly. (Note that since m > n log q, the number of
elements in the domain, 2m , is much larger than the number of elements in the range,
q n .)
We say collision x, x′ ∈ {0, 1}m when Ax = Ax′ .

• A = (a⃗1 , . . . , a⃗m ) ∈ Zn×m

q defines a q-ary lattice

L⊥ (A) = {z ∈ Zm : Az = 0 mod q}.

• Hence, SIS is SVP on L⊥ (A).

9
 • A syndrome u ∈ Znq defines a coset
L⊥
u (A) = {x ∈ Z : Ax = u mod q}
m

of L⊥ (A).
Remark 2.1.1. We are assuming that A has n-linearly independent columns. Hence,
A : Zm → Znq is onto, so |Zm /L⊥ (A)| = q n , i.e., det L⊥ (A) = q n .

Worst-Case / Average-Case reduction

Finding a short nonzero z ∈ L⊥ (A) for uniformly random A ∈ Zn×mq , where m ≈ n ln q
⇒ Solving GapSV Pβ √n , SIV Pβ √n on any n-dimensional lattice.

Algorithm for reduction

Repeat m-times.
Pick a random lattice point vi ∈ L, where L is an n-dimensional lattice.
Gaussian sample a point in 1q L around the lattice point vi ∈ L.
Hence, each sampled point can be written as vi + 1q B⃗ ai , where 1q B⃗
ai is short.
Give the m Znq samples a⃗1 , . . . , a⃗m to SIS oracle. Note that
A = (⃗a1 , . . . , ⃗an ) ∈ Zqn×m
is uniform. We subdivided the sides of the given lattice by “q”. So each lattice domain
of L has q n subpoints inside.

SIS oracle outputs z1 , . . . , zm ∈ {−1, 0, 1} such that

z1 a⃗1 + · · · + zm a⃗m = 0 (mod q). (2.4)
∑
Therefore, zi (vi + 1q B⃗
ai ) is a lattice point of the given lattice L. Hence,
1
B(z1 a⃗1 + · · · + zm a⃗m ) (2.5)
q
is a short lattice vector in L since it is the sum of short vectors 1q B⃗
ai .

2.1.2 Learning With Errors

Learning With Errors (LWE)
• Search LWE: Find s ∈ Znq given noisy random inner products
a⃗1 ← Znq , b1 = ⟨s, a⃗1 ⟩ + e1 (2.6)
a⃗2 ← Znq , b2 = ⟨s, a⃗2 ⟩ + e2 (2.7)
..
.
√
where ei ← χ, χ Gaussian over Z with width αq. (αq > n)
A∈Zn×m
Znq × Zm
q
/ Zm
(s,e)7→st A+e q

• Decision LWE: Distinguish (A, bt = st A + et ) from uniform (A, bt ), where A =

(a⃗1 , . . . , a⃗m ).
Note that Search LWE ⇔ Decision LWE.

10
Lattice interpretation of LWE
L(A) := {z ∈ Zm : zt = st A mod q for some s ∈ Znq } = π −1 (im A)
Zm
π

Znq
A / Zm
s7→st A q

Then, LWE ⇔ BDD on L(A). (Remark: From zt = st A + e, we obtain zt by BDD, then

solve the simultaneous equation zt = st A mod q to obtain s.)

SIS versus LWE

• Regev: LWE ≥ GapSVP, SIVP quantumly. (Peikert et al. showed LWE ≥ GapSVP
classically. But, classical reduction LWE ≥ SVP, or LWE ≥ SIVP is unknown.)

• SIS ≥ LWE:
If we find short z such that Az = 0, then from bt = st A + et , we find bt z = 0 + et z;
if (A, bt ) is LWE, then bt Z is short; if (A, bt ) is not LWE, then bt z rather well
spread.

• SIS ≤ LWE quantumly.

Simple properties of LWE

1. Easy to check a candidate solution s′ ∈ Znq : test if b − ⟨s′ , ⃗a⟩ is small.
If s ̸= s′ , then b − ⟨s′ , ⃗a⟩ = ⟨s − s′ , ⃗a⟩ + e is well spread in Zq .

2. Shift the secret by any t ∈ Znq ,

(⃗a, b = ⟨s, ⃗a⟩ + e) → (⃗a, b′ = b + ⟨t, ⃗a⟩ = ⟨s + t, ⃗a⟩ + e). (2.8)

random ts → random self-reduction.

We obtain many new LWEs with essentially the same solutions. Hence, we can
boost success probabilities arbitrarily close to 1.

Proof of equivalence of Search/Decision of LWE.

Suppose that D solves decision-LWE, i.e., it perfectly distinguish between (⃗a, b = ⟨s, ⃗a⟩+e)
and uniform (⃗a, b). We want to solve search LWE; i.e., given pairs (⃗a, b), find s. To find
s1 ∈ Zq , it suffices to test whether s1 = 0 because we can shift s1 by 0, 1, . . . , q − 1, i.e.,
choose t = (0, 0, . . . , 0) or t = (1, 0, . . . , 0), t = (2, 0, . . . , 0), . . . , t = (q − 1, 0, . . . , 0). For
each (⃗a, b), choose r ← Zq . Invoke D on pairs (a⃗′ = ⃗a − (r, 0, . . . , 0), b). Since

b = ⟨s, ⃗a⟩ + e
⟨ ⟩
= s, a⃗′ + (r, 0, . . . , 0) + e
⟨ ⟩
= s, a⃗′ + s1 r + e,
⟨ ⟩
we see that if s1 = 0, then b = s, a + e is LWE, and if s1 ̸= 0, then b is uniform.
⃗′

11
Decision-LWE with ‘Short’ Secret
We may assume that the secret is short, i.e., drawn from the error distribution χn . In
this case, we say that our LWE is in Hermite Normal Form (HNF of LWE).

1. Draw samples to get (Ā, b̄t = st Ā + ēt ) for square invertible Ā.

2. Transform additional samples of LWE (⃗a, b = ⟨s, ⃗a⟩ + e) to a⃗′ = −Ā−1⃗a,

⟨ ⟩
′ ⃗′
b = b + b̄, a
⟨ ⟩
= ⟨s, ⃗a⟩ + e + Āt s + ē, a⃗′
⟨ ⟩ ⟨ ⟩
= ⟨s, ⃗a⟩ + Āt s, a⃗′ + ē, a⃗′ + e
⟨ −1
⟩ ⟨ ⟩
= ⟨s, ⃗a⟩ + s, Ā(−Ā )⃗a + ē, a + e
⃗′
⟨ ⟩
= ē, a⃗′ + e.

(a⃗′ , b′ ) is LWE with secret ē. Then we obtain s from b̄t = st Ā + ēt .

2.2 Cryptosystems
2.2.1 Public-Key Cryptosystem using LWE
(Due to Regev)
A ← Zn×m
q (i.e., uniformly random n × m matrix over Zq ) open public.
s ← Zq Alice secret.
n

Public key of Alice

bt = st A + et . (2.9)
x ← {0, 1}m Bob secret.
Bob sends to Alice

u = Ax, (2.10)
u′ = bt x + bit · q/2. (2.11)

Alice decodes u′ − st u ≈ bit · q/2.

Note that (A, bt ) is LWE and (u, u′ ) uniformly random by left-over hash lemma when
m ≥ n log q.

2.2.2 Dual cryptosystem

A ← Zn×m
q open public as before.
Alice chooses a secret x ← {0, 1}m .
Alice’s public key
u = Ax. (2.12)

12
(by LHL, uniform if m ≥ n log q.)
Bob chooses a secret s ← Znq .
Bob sends

bt = st A + et , (2.13)
b′ = st u + e′ + bit · q/2. (2.14)

Adding bit · q/2 does not change the uniformity of st A + et .

Alice decodes b′ − bt x ≈ bit · q/2.
Note that (A, u; b, b′ ) is a LWE pair.

2.2.3 More efficient Cryptosystem

A ← Zn×n
q open public.
Alice chooses a secret s ← χn and an error e ← χn .
Alice’s public key
ut = st A + et . (2.15)
Bob chooses a secret r ← χn , x ← χ, and x, x′ ∈ χ.
Bob sends

b = Ar + x (2.16)
b′ = ut r + x′ + bit · q/2. (2.17)

Alice decodes b′ − st b ≈ bit · q/2.

Note that (A, u; b, b′ ) is a Hermite normal form of LWE.

13
14
Chapter 3

Discrete Gaussians and Applications

3.1 Discrete Gaussians and sampling

3.1.1 Discrete Gaussians
Gaussian sampling
Define ( )
π∥x∥2
ρs (x) := exp − 2 . (3.1)
s
Note that ρs is rather flat if s is large, and steep if s is small.
Note that ∫
ρs (x)dx = sn . (3.2)
x∈R
n

Hence, vs := sρns is an n-dimensional Gaussian probability density. We define Fourier

Transform as ∫
ĥ(w) = h(x)e−2πi⟨x,w⟩ dx. (3.3)
R n

Hence,
∫
ρ̂s (y) = ρs (x)e−2πix·y dx (3.4)
R n

∫ (
∥x∥2
)
−π +2ix·y
s2
= e dx (3.5)
∫R
n

∑ x 2
−π i ( si +iyi s) −π(∥y∥s)2
= e e dx (3.6)
Rn
= sn ρ 1 (y). (3.7)
s

Hence, if ρs (x) rather steep, then ρˆs is rather flat, and vice versa.

Remark 3.1.1.
∫
e−πx dx = 1,
2
(3.8)
∫ R
x 2
e−π( s ) dx = s. (3.9)
R

15
Poisson summation formula
Let

f (x) : R → C (3.10)
∑
F (θ) := f (θ + n) : S 1 → C, (3.11)
n∈Z

where S 1 = [0, 1]/0 ∼ 1. Then the Fourier series of F (θ) is

∑
F (θ) = an e2πinθ , (3.12)
n∈Z

where
∫ 1
an = F (θ)e−2πinθ dθ (3.13)
∫ 1 (∑ )
0

= f (θ + k) e−2πinθ dθ (3.14)
0
∫ ∞
k

= f (θ)e−2πinθ dθ (3.15)
−∞

= fˆ(n), (3.16)
∑ˆ
i.e., F (θ) = f (n)e2πinθ .
In particular, we obtain Poisson Summation Formula
∑ ∑
F (0) = f (n) = fˆ(n). (3.17)
n∈Z n∈Z

In general, for h : Rn → C,
ĥ(Zn ) = h(Zn ). (3.18)

Generalized Poisson summation formula

Let f : Rn → C.
f (L) = det L∗ fˆ(L∗ ), (3.19)
where L ⊂ Zn is a lattice and

L∗ = {x ∈ Rn : x · y ∈ Z, ∀y ∈ L} (3.20)

is called the dual lattice of L.

Proof. Let L = AZn for some n × n matrix A.

f (L) = (f ◦ A)(Zn ) (3.21)

\
= (f ◦ A)(Zn ) (3.22)

16
by Poisson summation formula. Let’s compute
∫
f[◦ A(y) = e−2πi⟨x,y⟩ (f ◦ A)(x)dx (3.23)
R n

putting Ax =: x′
∫
1
e−2πi⟨A x ,y⟩ f (x′ )dx′
−1 ′
= (3.24)
det A R n
∫
1 −2πi⟨x′ ,A−1 y⟩
T
= e f (x′ )dx′ (3.25)
det A R n

1
= · fˆ(A−T y). (3.26)
det A
Hence,
1 ˆ −T n
f[
◦ A(Zn ) = f (A Z ) (3.27)
det A
= det L∗ fˆ(L∗ ), (3.28)

because in general,

if L = L(B), B = (b1 , . . . , bn ), (3.29)

then L∗ = L(D), D = (d1 , . . . , dn ), (3.30)

where bi · dj = δij , i.e.,

BT D = I, (3.31)
i.e., L∗ = L(B−T Zn ). Note that det L∗ = (det L)−1 .
Corollary 3.1.2. ρr (L + c) ∈ rn det L∗ (1 ± ε) if r ≥ ηε (L), i.e., |ρ 1 (L∗ \ 0)| ≤ ε, i.e., ρ 1
r r
is very steep.
Proof.
∑
ρr (L + c) = ρr (x + c) (3.32)
x∈L
∑
= ρr,−c (x) (3.33)
x∈L
∑
= det L∗ ρd
r,−c (y) (3.34)
y∈L∗
∑
= rn det L∗ e2πi⟨c,y⟩ ρ 1 (y) (3.35)
r
y∈L∗
= rn det L∗ (1 ± ε). (3.36)

Smoothing parameter [MR04]

ηε (L), defined above, is called the smoothing parameter, because if r ≥ ηε (L), then ρr is
rather flat, smooth, and ρr (L+c) is almost uniform with respect to c. More quantitatively.
√
• ηε (L) ≥ n/λ1 (L∗ ) where ε = 2−n (Micciancio and Regev [MR04]).

17
 [ ]
• ∃ε ≤ 2e−πs such that ρs (c + Z) ∈ 1 ± s for all c ∈ R. Just we compute (note
2 ε
1−ε
that ρs (c + Z) ≤ ρs (Z))
∑
∞
2e−πs ε
2

e−π(sn) <
2
2 < (3.37)
n=1
1 − e−πs 2
1−ε

for some ε < 2e−πs . (True if ε is sufficiently close to 2e−πs .)

2 2

• The above example can be generalized to lattice L ⊂ Zn ⊂ Rn .

∃ε < 2ne−π( M ) such that ρs (c + L) ∈ [1 ± ε]sn for all c ∈ Rn , where M = max ∥bei ∥.
s 2

√ i
Especially if s > log nM , then ρs (c + L) ∈ (1 ± ε) poly(n)
1
.
Remark 3.1.3.
π∥x∥2
• ρs (x) = e− = ρs (x1 ) · · · ρs (xn )
s2

∏ n ( )
e < ε i
• ρs (L(B)) ≤ ρs (L(B)) 1+ sn for some ε1 , . . . , εn , where
i=1
1 − εi
 ( )2 
s
εi < 2 exp −π . (3.38)
∥Bei ∥

e follows from
ρs (L(B)) ≤ ρs (L(B))
 
f1 ∥
∥b ∗
 .. 
B = Q . , (3.39)
0 fn ∥
∥b
where Q is orthogonal.

Discrete Gaussians
Definition 1. Discrete Gaussian distribution over coset c + L is defined as
ρs (x)
Dc+L,s (x) = (3.40)
ρs (c + L)
for all x ∈ c + L.
Note that if s is sufficiently large √
(e.g., s > ηε (L)), then the denominator is very close
to sn det L∗ (e.g., with ε = 2−n , s > n/λ1 (L∗ )), and the numerator is the restriction of
ρs (x) on c + L. Hence, we only√ obtain∗ exponentially small information about c + L when
sampled from Dc+L,s if s ∼ n/λ1 (L ).

Choose x ∈ Zn from DZn ,s , where s > ηε (L). Reveal the coset x + L. Then every
coset c + L is almost equally likely, i.e., the distribution is almost uniform over Zn /L.
Given x ∈ c + L, it has the conditional distribution Dc+L,s .
Let
A ← Zn×m
q , i.e., uniformly (3.41)
x ← DZm ,s (3.42)

18
define fA (x) := Ax(= u) ∈ Znq . Then, inverting fA ⇔ decoding uniform syndrome u ⇔
solving SIS for A. (Solving Ax = u is equivalent to solving [A|u] [ −1x ] = 0.)
Conditional distribution when Ax = u is DL⊥u (A),s , where

L⊥
u = {x ∈ Z |Ax = u mod q}.
n

3.1.2 Sampling
Algorithms of Gaussian Sampling of DL⊥u (A),s
(As remarked before, DL⊥u (A),s sample does not reveal syndrome u if
√
log m max ∥bei ∥ ≤ s,

where S = (b1 , . . . , bm ) is a short enough basis of L⊥ (A), since εi = 1

O(poly(m))
in this
case.)

Nearest plane algorithm with randomized rounding

Gaussian sample a hyperplane in the coset L⊥ u (A) which is parallel to span{s1 , . . . , sm−1 },
⊥
where S = {s1 , . . . , sm } is a basis of L (A). Then consider the Z span of s1 , . . . , sm−1
displaced by the closest vector from the origin to the chosen hyperplane. Do Gaussian
sampling on∩ this displaced (m − 1)-dimensional lattice. Iterate this process. Note that
ρs ((c + L) plane) depends only on dist(0, plane), since ρs (x) depends only on ∥x∥.

Remark 3.1.4. Gaussian nearest plane algorithm for sampling DL⊥u (A),s is not efficient
and inherently sequential. We need a more efficient Gaussian sampling.

Randomized Babai’s roundoff algorithm. [Bab85]

Babai’s roundoff algorithm for finding the representative of a coset in the fundamental
parallelepiped can be written as

c 7→ Sf rac(S −1 c), (3.43)

where S = {s1 , . . . , sm } is a basis of L⊥ (A).

[ )m
Naive randomized rounding: Note that f rac(S −1 c) ∈ − 21 , 12 ⊂ Rm . Gaussian sam-
ple from f rac(S −1 c) + Zm , i.e., instead of the deterministic f rac(S −1 c), we could get
f rac(S −1 c) + ⃗p for some ⃗p ∈ Zm . Then apply S to obtain x. But then we have non-
spherical Gaussian distribution of x, because even though f rac(S −1 c) + ⃗p is spherical
Gaussian, when we apply S to f rac(S −1 c) + ⃗p to obtain x, we have nonspherical discrete
Gaussian such that
Ex (xxt ) ≈ S · S t , (3.44)
where S = (s1 , . . . , sm ) is a short basis of L⊥ (A), i.e., it leaks some information about
short basis S.

Breakthrough: Gaussian correction

19
Note that the sum of the Gaussian distribution is again Gaussian with the sum of the
covariances as its covariance. (The probability distribution of the sum of two random
variables X1 and X2 is
∫
PX1 +X2 (y) = PX1 (x)PX2 (y − x)dx.

Hence, P̂X1 +X2 = PbX1 P̂X2 . In particular, if PX1 and PX2 are Gaussian with covariances
s21 and s22 , respectively, then PX1 +X2 is Gaussian with covariance s21 + s22 .)
∑ ∑ ∑
1. Generate perturbation p with covariance 2 = σ 2 I − 1 , where 1 = SS t , and
σ > s1 (S), the largest singular value of S.
2. Randomly round off c + p to obtain a random sample
S · f rac(S −1 (c + p)) + L⊥ (A).

3. Then add −p.

3.2 Applications
3.2.1 Identity Based Encryption
Identity Based Encryption
• A: n × m matrix, master public key.
• u = H(Alice): hashed identity of Alice, public.
Master finds a Gaussian short element in fA−1 (u), i.e., x ← fA−1 (u) (Master has a short
basis of L⊥ (A)), and give Alice x as her secret key.

I want to send a message bit to Alice so that only Alice can decode. Choose Gaussian
short s, e ∈ Znq , e′ ∈ Zq
bt := st A + et (3.45)
q
b′ = st u + e′ + bit · (3.46)
2
Alice decodes: b′ − bt x ≈ bit · 2q .
(Note that this protocol is just a little modification of dual LWE cryptosystem.)
It seems that it is required to have a lattice together with a short basis when we apply
SIS or LWE to cryptography. But it is not a simple job to generate a lattice together
with a short basis.
The following signature protocol is a typical application of a lattice together with a
short basis.
• pk = A, sk = short basis of L⊥ (A).
• H : {0, 1}∗ → Znq random oracle.

• sign(msg): let u = H(msg), and output Gaussian x ← fA−1 (u).

• verif y(msg, x): check fA (x) = Ax = H(msg) and x is short enough.

20
Chapter 4

Constructing Trapdoors and More

Applications

4.1 Strong trapdoor generation and inversion algo-

rithms
4.1.1 Methods
Step 1: Gadget G and Inversion Algorithms
Let q = 2k (It could be generalized to an arbitrary q). Define a 1 × k matrix g :=
[1 2 4 · · · 2k−1 ] ∈ Z1×k
q . Then the columns of
 
2
 −1 2 
 
 −1 
 
S= ..  (4.1)
 . 
 
 2 
−1 2
is a lattice basis of L⊥ (g), and Se = 2Ik (Gram-Schmidt orthogonalization).
For q not a power of 2, let k = ⌈log(q)⌉ so q < 2k . Let g = [1, 2, . . . , 2k−1 ] ∈ Z1×k
q as
before. Then the columns of
 
2 q0
 −1 2 q1 
 
 −1 q2 
 
S= ..  ∈ Z
k×k
. (4.2)
 . . . 
 
 2 qk−2 
−1 qk−1
∑
k−1
⊥
is a short basis of L (g), where q = 2i qi is the binary expansion of q. (Note that
i=0
det S = q.)
• Inversion of LWE with only one equation

gq = sg + e = [s + e0 , 2s + e1 , . . . , 2k−1 s + ek−1 ] mod q, (4.3)

21
 where s ∈ Zq , small ei ∈ Z.
Get least significant bit from 2k−1 s + ek−1 , i.e., write s = s0 + s1 2 + · · · + sk−1 2k−1 ,
then
2k−1 s + ek−1 mod q = 2k−1 s0 + ek−1 . (4.4)
Hence, s0 = 0 if 2k−1 s + ek−1 is short and s0 = 1 if 2k−1 s + ek−1 is not short. Then
consider 2nd to the last, i.e.

2k−2 s + ek−2 = 2k−2 (s0 + 2s1 ) + ek−1 mod q. (4.5)

We subtract 2k−1 s0 to obtain 2k−1 s1 + ek−2 . Then we

[ qgetq )s1 in the same way as
before. This method works exactly when every ei = − 4 , 4 .

• Inversion of fg (x) = ⟨g, x⟩ = u.

For i ← 0, 1, . . . , k − 1, choose xi ← (2Z + u) by Gaussian sampling. Let u ←
(u − xi )/2 ∈ Z. Details are as follows. Note

⟨g, x⟩ = x0 + 2x1 + 22 x2 + · · · + 2k−1 xk−1 = u. (4.6)

Hence, x0 is even or odd according to u. x0 ← 2Z + u, Gaussian sampling from

2Z + u. Once we get x0 , (u − x0 )/2 = x1 + 2x2 + · · · . The same method works to
find x1 , · · · .

Define  
···g···
 ···g··· 
 
G = In ⊗ g =  ...  ∈ Zqn×nk , (4.7)
 
···g···
where k =⌈ log q⌉ as before. Now fG−1 , gG
−1
reduce to n parallel calls to fg−1 , gg−1 .
Also applies to HG for any invertible H ∈ Zn×n
q by considering fG−1 ◦ H−1 or H−1 ◦ gG −1
.

Step 2: Randomize G to obtain uniformly random A

Consider n × (m̄ + nk) matrix [Ā|G] for uniform Ā ∈ Zqn×m̄ . Then it is easy to solve SIS
and LWE for [Ā|G].
( )
−1 x1
• SIS for [Ā|G] is f[Ā|G] (u) = x, where (Ā|G)x = u, X = , x1 ∈ Zm̄ , x2 ∈ Znk .
x2

Āx1 + Gx2 = u. (4.8)

To obtain such x, choose small x1 , then apply fG−1 to u − Āx1 to get x2 .

−1
• LWE for [Ā|G], st (Ā|G) + et = (st Ā + et1 |st G + et2 ) = (bt1 |bt2 ). Apply gG to bt2 to
obtain s, since
st G + et2 = bt2 . (4.9)
And confirm s satisfies st Ā + e1 = bt1 .

22
 To obtain random matrix A, choose short Gaussian R ← Zm̄×n⌈log q⌉ and
( )
I −R
A := (Ā|G) (4.10)
0 I
= (Ā|G − ĀR). (4.11)
A is uniform if ĀR is uniform. If m̄ ≈ n log q, ĀR is uniform, since R → ĀR is
n×n⌈log q⌉
uniform from Zm̄×n⌈log q⌉ → Z(q ) , and left over hash lemma applies if 2 ≈ q ,
m̄ n

i.e., m̄ ≈ n log q. (Note that 0 −R is unimodular, hence the above construction is

I R

just a base change.)

Now we have constructed uniformly random A = (Ā|G − ĀR).
( )
R
Definition 2. R is a trapdoor for A with tag H ∈ Zq , which is invertible, if A
n×n
=
I
HG.
Quality of R is
s1 (R) = max ∥Ru∥. (4.12)
∥u∥=1

(maximal singular value)

From random matrix theory, we know
√ √
s1 (R) ≈ ( #rows + #colums)r (4.13)
for Gaussian entries with standard deviation r.
Remark 4.1.1. Let S ∈ Zw×w be any basis for L⊥ (G). (w = nk)
A ∈ Zn×m
q have trapdoor R ∈ Z(m−w)×w with tag H ∈ Zqn×n . Then L⊥ (A) is generated
by the basis ( )( )
I R I 0
SA = , (4.14)
0 I W S
where W ∈ Zw×m̄ is an arbitrary solution to
GW = −H −1 A(I|0)T mod q. (4.15)
Note that both sides are n × m̄ matrices, and m = m̄ + w. Hence, (I|0) is an m̄ × m
matrix. Eq. (4.15) has many solutions since W has w × m̄ unknowns and the right hand
side gives n × m̄ conditions and G is a rank n matrix.. It is easy to check ASA = 0
mod q, and det(SA ) = det S = q n = det(L⊥ (A)). (We assume Zm ∋ x → Ax ∈ Znq is
onto.) Let us consider the Gram-Schmidt Orthogonalization of SA ,

SeA = T
g B, (4.16)
( ) ( ) ( )
I 0 I R e I 0 e = ∥S∥.
e
where B = and T = . B= , hence ∥B∥
W S 0 I 0 Se
g
Now we prove ∥T e
B∥ ≤ s1 (T )∥B∥.
Let
B = QDU, T B = Q′ D′ U ′ (4.17)
by the Gram-Schmidt decomposition of B and T B, respectively, where Q is orthogonal,
D is positive diagonal, and U is upper triangular.
T QDU = Q′ D′ U ′ ⇒ T ′ D = D′ U ′′ , (4.18)

23
where T ′ = Q′−1 T Q and U ′′ = U ′ U −1 . Then
g
∥T e = s1 (T )∥B∥
B∥ = ∥D′ ∥ ≤ ∥D′ U ′′ ∥ = ∥T ′ D∥ ≤ s1 (T ′ )∥B∥ e = s1 (T )∥S∥,
e (4.19)

since the ith row of D′ U ′′ has the norm at least d′i,i , the i-th diagonal of D′ .
Since ( ) ( )
I 0 0 R
T = +
0 I 0 0
and s1 (T ) ≤ s1 (R) + 1, it follows that

∥SeA ∥ ≤ (s1 (R) + 1)∥S∥.

e (4.20)

24
 (R)
Suppose that A I
= G.

• Given a LWE problem with coefficient matrix A,

bt = st A + et , (4.21)

we can recover s from the LWE problem with coefficient matrix G,

( ) ( )
t R t t R
b =s G+e . (4.22)
I I
( ) ( )
It works if each entry of et RI is in − 4q , 4q , i.e., ∥e∥ < 4s q R .
1( I )

• Sampling Gaussian preimage.

( )
Given u, sample z ← fG−1 (u) and output x = RI z ∈ fA−1 (u).
Then we have Ax = Gz = u as desired, i.e., we obtained an SIS solution x with
respect A from an SIS solution with respect to G.
( )
But there is the problem as before that RI z is nonspherical even though( z)is
spherical, i.e., it leaks R. This can be cured as before. The covariance of x = RI z
is (( ) ( )t )
∑ R R
= Ex (x · xt ) = Ez z · zt ≈ s2 RRt , (4.23)
I I
when z spherical Gaussian with deviation s.

∑
Choose s > s1 (R) and let 2 = s2 I − RRt >
∑0.
Generate perturbation p with covariance
( ) 2 . Sample a spherical z such that
Gz = u − Ap. Output x = p + RI z. This algorithm generates a spherical
discrete Gaussian over L⊥
u (A).

4.2 Applications
Efficient IBE
1. Choose A = (Ā| − ĀR). Let mpk = (A, u), msk = R (A has trapdoor R with tag
0).

2. map: id → invertible Hid ∈ Zn×n q

choose skid : x ← fA−1id (u) using the above algorithm, sampling Gaussian preimage,
where Aid = A + (0|Hid G) = (Ā|Hid G − ĀR).

3. Encrypt to Aid , decrypt using skid as in dual public key cryptosystem.

25
26
 Part II

Introduction to Ring-LWE

27
Chapter 5

Preliminaries for Ring-LWE

cryptography

5.1 Notations
In Part II, we use the notations in [LPR13].
∩ [ 1 1)
• ∀ā ∈ R/Z, JāK ∈ R denotes the unique representative, where a ∈ (ā + Z) −2, 2 .
∩
• ∀ā ∈ Zq , JāK denotes the unique representative a ∈ (ā + qZ) [−q/2, q/2).

• [k] = {0, 1, . . . , k − 1}.

• Zm = Z/mZ.

• Z∗m ⊂ Zm : the set of invertible elements mod m.

• |Z∗m | = φ(m): Euler totient.

∗
• H = {x ∈ CZm : xi = xm−i , ∀i ∈ Z∗m }. Note that if (i, m) = 1 (relatively prime),
then (m − i, m) = 1 also.

• H∼
= R[n] , n = φ(m).
( √ )
• B= √1
I √−1J unitary basis of H, where
2 J − −1I
   
1 1
 1   1 
   
I= .. , J = . ,
 .   .. 
1 1
√
+ em−i ) for i < m/2 and i ∈ Z∗m ,
√1 (ei √−1 (ei − em−i ) for i > m
and i ∈ Z∗m . We
2 2 2
read B as a Z∗m -by-[n] matrix.

29
5.2 Gaussians and Subgaussian Random Variables
We follow [LPR13] as before, giving some details.

Definition 3. Random variable X over R is δ-subgaussian with parameter s > 0 if for

all t ∈ R
E[exp(2πtX)] ≤ exp(δ) · exp(πs2 t2 ). (5.1)

Lemma 5.2.1. P r(|X| ≥ α) ≤ 2 exp(δ − πα2 /s2 ).

Proof. From the definition, Markov inequality1 says

P r(X > α) ≤ exp(δ) · exp(πs2 t2 − 2πtα) (5.2)

( )
πα2
for any t. RHS becomes minimum at t = s2 , and its value is exp(δ) · exp − s2 .
α

Similarly, P r(X < −α) ≤ exp(δ) · exp(πs2 t2 − 2πtα).

Example
√ 5.2.1. If E(X) = 0 and |X| ≤ B, then X is 0-subgaussian with parameter
B 2π.

Proof. Let p(x) be a probability distribution of random variable X. Then

∫ B
E(exp(2πtX)) = e2πtx p(x)dx, (5.3)
−B

∫B ∫B
where E(x) = −B xp(x)dx = 0 and −B p(x) = 1. Simplex method says that the max-
∫B
imum of −B exp(2πtx)p(x)dx occurs when p(x) is a boundary point of the simplex of
the probability space satisfying the given conditions, i.e., p(x) = (δB (x) + δ−B (x))/2, and
2πtB −2πtB
its value is e +e = cosh(2πtB) ≤ exp(2π 2 B 2 t2 ). Hence, X is 0-subgaussian with
√ 2
parameter B 2π.

Remark 5.2.2. We can prove ex + e−x ≤ 2ex

2 /2
by series expansion.

Lemma 5.2.3. If the conditional probability

∑ ∑P r(Xi |X1 , . . . , Xi−1 ) is δi -subgaussian
∑ 2with
parameter si for i = 1, . . . , k, then Xi is ( δi )-subgaussian with parameter ( si )1/2 .

Proof. We may assume k = 2.

E(exp 2πt(X1 + X2 )) = EX1 (exp(2πtX1 )EX2 (exp 2πtX2 |X1 )) (5.4)

≤ exp(δ1 + δ2 ) exp(π(s21 + s22 )t2 ). (5.5)

( )
Lemma 5.2.4. Let X be δ-subgaussian with parameter s. Then for any t ∈ 0, 2s12 ,
( )−1
1
E(exp(2πtX )) ≤ 1 + 2 exp(δ) ·
2
−1 . (5.6)
2ts2
∫∞ ∫∞
1
For any positive and increasing function f , E(f (X)) = −∞
f (x)p(x)dx ≥ α
f (x)p(x)dx ≥
f (α)P r(X > α), so P r(X > α) ≤ E(f (X))/f (α).

30
Proof. By Lemma 5.2.1,

exp(2πtr2 )P r(|X| > r) ≤ 2 exp(δ) exp(π(2t − 1/s2 )r2 ), (5.7)

and since t < 1/2s2 by assumption, 2t − 1/s2 < 0, so

lim exp(2πtr2 )P r(|X| > r) = 0. (5.8)

r→∞

Now let P r(|X| > r) = f (r). Then

df = −(p(x) + p(−x))dx, (5.9)

∫∞ ∫ −r
where p(x) is the probability density, since f (r) = r p(x)dx + −∞ p(x)dx. Hence,
∫ ∞
2
E(exp(2πtX )) =
2
e2πtr (p(r) + p(−r))dr (5.10)
0
∫ ∞
2
= − e2πtr d(P r(|X| ≥ r)) (5.11)
0
∫ ∞
2
= 1+ d(e2πtr )P r(|X| ≥ r) (5.12)
0
2
(Since e2πtr P r(|X| > r) = 1 when r = 0, and by (5.8).)
∫ ∞
= 1+ P r(|X| > r)4πtr exp(2πtr2 )dr (5.13)
0
∫ ∞
≤ 1 + 8πt exp(δ) r exp(−πr2 /s2 + 2πtr2 )dr (5.14)
0
(Since t > 0 and by (5.7).)
( )−1
1
= 1 + 2 exp(δ) −1 (5.15)
2ts2
( ( )−1 )
1
≤ exp 2 exp(δ) −1 . (5.16)
2ts2

Lemma 5.2.5. If X1 , . . . , Xk are random variables each of which is δ-subgaussian with

parameter s conditioned on any values of the previous ones, then for any r > k ′ s2 /π
where k ′ = 2k exp(δ), we have that
(∑ ) ( ( ( ))
′ πr )1/2 πr
Pr Xi > r ≤ exp k 2 ′ 2
2
− ′ 2 −1 . (5.17)
ks ks
Proof. From the previous lemma,
( ( )−1 )
( ( ∑ )) 1
E exp 2πt Xi2 ≤ exp 2k exp(δ) −1 , (5.18)
2ts2

where 0 < t < 1/2s2 . Hence,

( ( )−1 )
(∑ ) 1
Pr Xi2 > r ≤ exp 2k exp(δ) −1 − 2πtr . (5.19)
2ts2

31
Letting x = 2s2 t and A = πr/(s2 k ′ ) (note that 0 < x < 1 and A > 1 by assumption), the
expression inside the exponent can be written as
(( )−1 )
1
2k exp(δ) −1 − Ax . (5.20)
x
√
The minimum of ( x1 − 1)−1 − Ax is 2 A − A − 1, obtained at x = 1 − √1 .
A

Remark 5.2.6. Since 2α1/2 − α − 1 < −α/4 for all α ≥ 4,

∑ ( πr )
P r( Xi2 > r) ≤ exp − 2 (5.21)
4s
for any r ≥ 4k ′ s2 /π.

Definition 4. An Rn -valued random variable X is δ-subgaussian with parameter s if for

all unit vectors u ∈ Rn , ⟨X, u⟩ is δ-subgaussian with parameter s.

Note that if the coordinates of X are independent and all are δ-subgaussian with
parameter s, then X is nδ-subgaussian with the same parameter s. (If u = (u1 , . . . , un )
and u21 + · · · + u2n = 1, then ui Xi is δ-subgaussian with parameter |ui |s.)

Corollary 5.2.7. For i = 1, . . . , k, let Xi be random vectors in Rn , and Ai n × n

matrices. For δi , si ≥ 0, suppose that X∑ i is δi -subgaussian
∑ with parameter si condi-
tioned∑on any values of X1 , . . . , Xi−1 . Then Ai Xi is ( δi )-subgaussian with parameter
λmax ( s2i Ai ATi )1/2 .

Proof. For any unit vector u ∈ Rn ,

⟨ ⟩
∑ ∑ ∑⟨ ⟩
Ai Xi , u = ⟨Ai Xi , u⟩ = Xi , ATi u . (5.22)
i i i
⟨ ⟩
∑parameter si ∥Ai u∥2 conditioned on any value
Note that Xi , ATi u is δi -subgaussian with T

of the previous ones. Hence, the sum is ( δi )-subgaussian with parameter

(∑ )1/2 ∑
s2i ∥ATi u∥22 = ((uT s2i Ai ATi )u)1/2 , (5.23)
∑
whose maximum over all unit vectors u is λmax ( s2i Ai ATi )1/2 .

5.3 Lattice Background

∑
Let Λ = L(B) = { j zj bj : zj ∈ Z} be a lattice in H generated by a basis
∑ B = {bj } ⊂ H.
∨
We define dual lattice of Λ ⊂ H as Λ = {y ∈ H : ∀x ∈ Λ, ⟨x, ȳ⟩ = xi yi ∈ Z}. Note
that this is actually the complex conjugates of the dual lattice as usually defined in
C
⟨ . If ⟩Λ = L(B), where
n
B = {bj } ⊂ H, the dual basis D = {dj } is characterized by
bi , d̄k = δjk , i.e., BDT = I, i.e., D = B T −1 .

Remark 5.3.1.

32
 1. ρs
sn
is a probability distribution on Rn . Hence, ρssn|Λ det Λ is almost a probability
distribution on the lattice Λ. In particular, ρss(Λ)
n det Λ ≈ 1. More precisely, if
s ≥ ηε (Λ), where ηε (Λ) is the smoothing parameter defined earlier, then ρs (Λ + c) ∈
(1 + ε)sn det(Λ)−1 ([Reg05]).

2. For any n-dimensional lattice Λ and s > 0, a point sampled from DΛ,s = ρsρ(Λ)
s
has
√ −2n
the Euclidean norm of at most s n except with probability at most 2 ([Ban93]).

3. There is an efficient algorithm that samples to within negl(n) statistical

√ distance of
˜
DΛ+c,s given c ∈ H, a basis B of Λ, and a parameter s ≥ maxj ∥bj ∥ω( log n) ([GPV08]),
where we define the discrete Gaussian probability distribution over Λ + c as
ρs (x)
DΛ+c,s(x) = , ∀x ∈ Λ + c. (5.24)
ρ(Λ + c)
Note c is not the center of DΛ+c,s , since we did not translate DΛ by c. In comparison,
c + DΛ is a c-centered distribution. Rather, we restricted ρs on Λ + c. So the
maximum probability of DΛ+c,s occurs at the nearest point to the origin.

5.3.1 Decoding
Λ ⊂ H: a fixed lattice.
x ∈ H: an unknown short vector.
We are given t such that t = x mod Λ. The goal is to recover x.

First attempt
∑ ∑
A basis B = (b1 , . . . , bn ) is known, t = ci bi , and claim x = (ci − ⌈ci ⌋)bi , i.e.,
Babai’s round off algorithm with respect to the basis B [Bab85].
∑
Problem: If the basis B are not short, then (ci − ⌈ci ⌋)bi not short in general. Hence,
in that case it couldn’t be x, because x is rather short. This algorithm succeeds when
|x| ≤ d, where the ball of radius d is in P (B).

Second attempt
Choose {vi }, a fixed set of n linearly independent and typically short vectors in the dual
lattice Λ∨ ({vi } need not be a basis of Λ∨ ). Denote the dual basis of {vi } by {b′i }, and
let Λ′ ⊃ Λ be the super lattice generated by {b′i }. Given ∑an input t = x mod Λ, we
′ ′ ′
re-express
∑ t in mod Λ with respect to the basis {bi } as c b
i i , c i ∈ R/Z, and output
′
i Jci Kbi ∈ H (Note that ci = ⟨x, v̄i ⟩ mod 1). Hence, the∑ output is equal to x if and
only if all the coefficients ai = ⟨x, v̄i ⟩ in the expansion x = ai b′i are in [− 21 , 12 ). (Note
that in general, b′i ∈ Λ′ is small but not necessarily in Λ.) Hence, the second attempt
works when x ∈ P (B ′ ). In general, the radius of the ball enclosed in P (B ′ ) is larger
than the radius of the ball enclosed in P (B) with the given basis B = (b1 , · · · , bn ) even
though Λ′ ⊃ Λ, because of the choice of {vi }.
Example 5.3.1. Define a lattice by
∑
x = (x1 , . . . , xn ) ∈ Λ ⊂ Zn if xi = 0 mod 2.
i

33
Then
1
Λ∨ = Zn ∪ (Zn + (1, 1, . . . , 1)).
2
A basis of Λ is
{(1, 1, 0, . . . , 0), (1, 0, 1, 0, . . .), (1, 0, . . . , 0, 1), (2, 0, . . . , 0)},
and a basis of Λ∨ is
{(1, 0, . . . , 0), · · · , (0, . . . , 0, 1, 0), 21 (1, . . . , 1)}.
Since
{vi } = {(1, 0, . . . , 0), · · · , (0, . . . , 0, 1, 0), (0, . . . , 0, 1)},
Λ ⊃ L({vi }). But ∥vi ∥ = 1 for i = 1, . . . , n, and Λ′ = L({vi })∨ = Zn ⊃ Λ.
∨

Discretization
Input Λ = L(B) with a good basis B = {bi }, x ∈ H, c ∈ H.
The goal is to discretize x to a point y ∈ Λ + c written y ← ⌊x⌉Λ+c , so that y − x is
not too large. Hence, it suffices to find a relatively short offset vector f from the coset
Λ + c′ = Λ + (c − x) and output y = x + f . Note that ⌊z + x⌉Λ+c and z + ⌊x⌉Λ+c are
identically distributed for any z ∈ Λ if our algorithm depends only on the coset Λ + c′ ,
and not on the particular representative. In this case, it is called valid discretization.

Coordinate-wise randomized rounding:

∑
Given a coset Λ + c′ , represent c′ = ai bi mod Λ for some coefficient ai ∈ [0, 1), then
randomly and ∑ independently choose fi from {ai − 1, ai } to have zero expectation,
√ and
output f = fi bi ∈ Λ + c′ . Note√that fi is 0-subgaussian with parameter 2π, hence f
is 0-subgaussian with parameter 2πs1 (B), since |fi | ≤ 1 and
 
f1
 f2 
 
B  ..  = f ,
.
fn
∑
where B = (b1 , . . . , bn ). More directly, let√u be a unit vector. ⟨f , u⟩ = fi ⟨bi , u⟩, and
fi ⟨bi , u⟩ is 0-subgaussian with parameter 2π · |⟨bi , u⟩|. Hence, ⟨f , u⟩ is 0-subgaussian
∑ 1/2 √
with parameter ( 2π|⟨bi , u⟩|2 ) ≤ 2πs1 (B).

5.4 Algebraic Number Theory Background

For a positive integer m, the mth cyclotomic number field is a field extension K = Q(ζm )
obtained by adjoining an element ζm of order m (primitive mth root of unity) to the
rationals. (Hence, Q(ζm ) = Q[ζm ].) The minimal polynomial of ζm is the mth cyclotomic
polynomial ∏
Φm (X) = (X − ωmi
) ∈ Z[X], (5.25)
i∈Z∗m

where ωm = e2πi/m .
Since n = |Z∗m | = φ(m) := degree of Φm , we can view K as a vector space of dimension
n over Q, which has a basis (ζm
j n−1
)j∈[n] = (1, ζm , . . . , ζm ), called the power basis.

34
 ∏
Remark 5.4.1. X m − 1 = d|m Φd (X), where d runs over all the positive divisors of m,
because an mth root of unity is a primitive dth root of unity for some divisor d of m, and
conversely a primitive dth root of unity is an mth root of unity if d divides m. (Another
remark: Decompose {0, 1, 2, . . . . , m − 1} according to gcd(j, m).) In particular,

Φp (X) = 1 + X + X 2 + · · · + X p−1

for any prime p, and by induction,

Xp − 1 tp − 1
r

Φpr (X) = pr−1 = = 1 + t + · · · + tp−1 , (5.26)

X −1 t−1
r−1
where t = X p . In general, for any m,

Φm (X) = Φrad(m) (X m/ rad(m) ),

where rad(m) is the product of all distinct primes dividing m. If m′ divides m, we can
m/m′
view K ′ = Q(ζm′ ) as a subfield of K = Q(ζm ) by identifying ζm′ with ζm . In general
Φpq (X) is not of simple form for distinct primes p and q, even though

φ(pq) = φ(p)φ(q) = (p − 1)(q − 1).

5.4.1 A key fact from algebraic number theory

Let m = Πl ml be a prime power factorization i.e., ml are powers of distinct primes. Then

Q(ζm ) ∼
= ⊗Kl ,
⊗ ∏
where Kl = Q(ζml ), via the correspondence l al ↔ l al , where on the right we embed
each al ∈ Kl into K as a subfield.

5.4.2 Canonical Embedding and Geometry

Let K = Q(ζm ), and ωm ∈ C a fixed primitive mth root of unity, for example e2πi/m . For
each i ∈ Z∗m , let
σi : K → C, ζm →
7 ωmi
.
Clearly σi = σ̄m−i , ∀i ∈ Z∗m , because σi σm−i = 1, i.e., σi = (σm−i )−1 = σ̄m−i . We define
the canonical embedding
∗
σ : K → CZm , a 7→ σ(a) = (σi (a))i∈Z∗m .
∗
Hence, σ(K) ⊂ H ⊂ CZm , where H is defined as before. Note that σ is a ring homomor-
phism from K to H, where multiplication and addition in H are both componentwise.
√
For a ∈ K, define ∥a∥2 = ∥σ(a)∥2 and ∥a∥∞ = maxi |σi (a)|. Then ∥ζ∥2 = n and
∥ζ∥∞ = 1.
The map ∑
Tr : K → Q, a 7→ σi (a)
i∈Z∗m

is called the trace. Note that

∑ ⟨ ⟩
Tr(a · b) = σi (a)σi (b) = σ(a), σ(b) . (5.27)
i

35
 The map ∏
N : K → Q, a 7→ σi (a)
i∈Z∗m

is called the norm.

Note that Tr(a) and N(a) can also be thought of as the trace and the determinant of
∗ ∗
the multiplication map σ(a) : CZm → CZm .
It is trivial to see that
N(ab) = N (a)N (b). (5.28)
⊗
With the canonical isomorphism K ∼ = Kl , we have
l

σ (⊗l al ) = ⊗l σ (l) (al ), (5.29)

∏
TrK/Q (⊗l al ) = TrKl /Q (al ), (5.30)
l
∏
N(a1 ⊗ · · · ⊗ ak ) = N(al )m/ml . (5.31)
l

5.4.3 The Ring of Integers and Its Ideals

Let R ⊂ K denote the set of all algebraic integers in a number field K, i.e., a ∈ R ⊂ K if
and only if it satisfies a monic integral polynomial. R is called the ring of integers. Note
that Tr, N : R → Z, and for cyclotomic number field K = Q(ζm ), R = Z[ζm ] ∼ Z[x]
= Φm (x) .
⊗
Hence, the power basis {ζm j
}j∈[n] is also a Z-basis of R. We can view R ∼
= l Rl as before.
Definition 5. Discriminant ∆K of K is ∆K = det(σ(R))2 .
 n
 
 m 
∆K =  ∏ 1  ≤ n ,
n
(5.32)
 p p−1 
primep|m

for the mth cyclotomic number field and n =√φ(m). ∆K ≤ nn follows from σ(R) =
1
span{σ(1), σ(ζm n−1
), . . . , σ(ζm )} and ∥σ(ζm
i
)∥ = n. Note that
∆K = | det(σi (ζm
j
))|2 (5.33)
= | det(Tr(ζm i j
ζm ))|, (5.34)
because ∑ ∑
Tr(xi xj ) = σk (xi xj ) = σk (xi )σk (xj ) = H T H,
k k
i
where xi = ζm and H = (σi (xj )).
I ⊂ K is called a fractional ideal if ∃d ∈ R such that dI ⊂ R is an integral ideal. It
is principal if I = uR for some u ∈ K. σ(I) ⊂ H called an ideal lattice. For an I ⊂ R,
define the norm as N (I) = |R/I| (= the number of cosets of I in R).
Note the following:
• Consider the lattices σ(R) ⊃ σ(I). Then N (I) = |σ(R)/σ(I)|, and σ(R) is the Z-
1 n−1 1 n−1
span of σ(1), σ(ζm ), . . . , σ(ζm ). σ(⟨a⟩) is spanned by σ(a), σ(aζm ), . . . , σ(aζm ).
i i
The j-th coordinate σj (aζm ) = σj (a)σj (ζm ) is stretched by σj (a). Hence, N (⟨a⟩) =
|N (a)|.

36
 • N (aI) = N (I)N (⟨a⟩) because |R/aI| = | RI ∥ aII | = N (I)N (⟨a⟩) and | aII | = | aR
R
|.

• N (IJ) = N (J)N (I)

(Case 1) I, J coprime
The Chinese remainder theorem says that R → R/I ⊕ R/J is onto, and its kernel is
I ∩ J = IJ. (It is trivial to see IJ ⊂ I ∩ J. To show that I ∩ J ⊂ IJ, let y ∈ I ∩ J;
then y = y · 1 = y(a + b) = ya(∈ IJ) + yb(∈ IJ), where a ∈ I and b ∈ J, since
I + J = R.) Remember that any element of IJ is of the form

a1 b1 + · · · + al bl for ai ∈ I, bj ∈ J. (5.35)

(Case 2) I = pm , J = pk for some prime ideal p.

Just note that R/p ≈ pn /pn+1 . (Since R is Dedekind, pn /pn+1 is singly generated;
the isomorphism is given by multiplying the inverse of the generator.)
(Case 3) The general situation can be reduced to either (Case 1) or (Case 2).

The norm of a fractional ideal I is defined by N (I) := N (dI)/|N (d)|, where

√ d∈R
is such that√dI ⊆ R. This is well-defined. Note that det(σ(I)) = N (I) ∆K , since
∗
det σ(R) = ∆K . (det σ(I) is the determinant of the lattice σ(I) ⊂ H ⊂ CZm .)

Lemma 5.4.2. √
√ 1/n √ 1/n 1/n
nN (I) < λ1 (I) ≤ nN (I) ∆K (5.36)

Proof. The upper bound is just Minkowski’s inequality. To prove the lower bound, let
v ∈ I such that ∥v∥ = λ1 (I). Since ⟨v⟩ ⊂ I,
 
∏ 
 
N(I) ≤ |N (v)| =  σi (v) .
 
i
∑
Note that ∥v∥2 = |σi (v)|2 by the definition of the metric on H. Also,
( ) n1
∏ 1∑
|σi (v)|2 ≤ |σi (v)|2 . (5.37)
i
n i
√ 1
Hence, nN (I) n ≤ ∥v∥.

Remark 5.4.3. It follows that GapSV Pγ of ideal lattice is trivial if γ = poly(n).

5.4.4 Duality
For more details, see [Con09].

Definition 6. For a fractional ideal I in K, its dual is defined as

I ∨ = {a ∈ K : Tr(aI) ⊆ Z}. (5.38)

Then σ(T ∨ ) is a dual lattice (more precisely, a conjugate dual lattice) of σ(I), because
the inner product in H is defined by Tr.

Definition 7. For any Q-basis B = {bj } of K, define a dual basis B ∨ = {b∨j }, where
Tr(bi b∨j ) = δij .

37
 Note that R∨ ⊃ R from the definition of integral elements, because Tr(r) ∈ Z for all
r ∈ R.

Lemma 5.4.4. I ∨ = I −1 R∨ (R∨ is called the codifferent, (R∨ )−1 the different).

Here I −1 := {x ∈ K : xI ⊂ R}. It is a fractional ideal.

Proof.

1) I −1 R∨ ⊂ I ∨ : trivial from the definitions of I −1 , R∨ and I ∨ .

2) Note that
 
R
∨
N(I ) =   (5.39)
I∨ 
 ∨ −1
I 
=   (5.40)
R
 ∨ −1  ∨ −1
 I  R 
=     (5.41)
 R∨   R 
 −1
R
=   ∆−1 (5.42)
I  K

= N (I)−1 ∆−1
K , (5.43)
∨
because I ⊂ R ⊂ R∨ ⊂ I ∨ , | JI | = det σ(J)
, | JI ∨ | = | JI |−1 , det I ∨ = (det I)−1 , and
√ det σ(I)
det(σ(R)) = ∆K . Note also

N(I −1 R∨ ) = N (I −1 )N (R∨ ) = N (I)−1 ∆−1

K , (5.44)

because N (IJ) = N (I)N (J) holds for general fractional ideals I, J ⊂ K. Hence,
I ∨ = I −1 R∨ .

Lemma 5.4.5. Let m be a power of prime p, m′ = m/p, and j an integer. Then


 φ(p)m′ if j = 0 mod m,
′
j
Tr(ζm ) = −m if j = 0 mod m′ , j ̸= 0 mod m, (5.45)

0 otherwise.

Proof. Let d = gcd(j, m), m̃ = m/d. Then

∑ ∑
j j α dj ′ α
Tr(ζm )= (ζm ) = (ζm ) , (5.46)
α∈Z∗m α∈Z∗m

where j = dj ′ . If α = α′ mod (m/d = m̃), dj ′ α′ = dj ′ (α + m

d
k) = dj ′ α + j ′ mk. Hence,
dj ′ α dj ′ α′
ζm = ζm . Therefore we have
j j/d
Tr(ζm ) = d TrQ(ζm̃ ) (ζm̃ ), (5.47)

since we have d such α′ s, and ζm

d
= ζm̃ .

38
 Note that Z∗m → Z∗m̃ is d-fold onto map when m = dm̃. Also note that
∑ {
i −1 if m = p,
ωm = (5.48)
∗
0 if m = pk , k ≥ 2,
i∈Zm

where ωm is a primitive mth root of unity, because

Φp (x) = 1 + x + · · · + xp−1 (5.49)

′ ′
Φm (x) = 1 + xm + · · · + xm (p−1) (5.50)

where m′ = pk−1 . The lemma follows, because ζm̃ is a primitive m̃-th root of unity.
j/d

Lemma 5.4.6.
⟨ g ⟩Let m be a power of a prime p, m′ = m/p, and let g = 1−ζp ∈ R = Z[ζm ].
Then R = m , p/g ∈ R, and ⟨g⟩ and ⟨p ⟩ are coprime for every prime integer p′ ̸= p.
∨ ′

Proof. We first show that g/m ∈ R∨ . It suffices to show that Tr(ζm

j
g/m) is an integer
for every j ∈ [φ(m)]. Note that
′
j
ζm j
g/m = (ζm − ζm
j+m
)/m. (5.51)


 ′
(ϕ(p) + 1)m (= m) if j = 0 mod m,
′
j
Tr(ζm − ζm
j+m
) = (−m′ ) − (−m′ ) = 0 if j = 0 mod m′ and j ̸= 0 mod m,


0 otherwise.

Note that in the second case, j ∈ [φ(m)], i.e., j = 0, · · · , m′ (p − 1) − 1, hence not only j
but also j + m satisfies j = 0 mod m′ and j ̸= 0 mod m.
We therefore have {
j 1 for j = 0,
Tr(ζm g/m) = (5.52)
0 otherwise.
To show R∨ = ⟨g/m⟩, we compute N (R∨ ) and N (g/m). Let m = pl .

N(R∨ ) = ∆−1
K (by Eq. (5.44)) (5.53)
( 1 )pl−1 (p−1) l−1
p p−1 pp pm/p
= = = (5.54)
pl mφ(m) mφ(m)
(5.55)
φ(m)
N(m) = m (5.56)
(5.57)
N(g) = N(1 − ζp ) = [NQ(ζp )/Q (1 − ζp )]m/p (5.58)
= [(1 − ζp )(1 − ζp2 ) · · · (1 − ζpp−1 )]m/p (5.59)
m/p
= p (5.60)

since
Φp (x) = (x − ζp ) · · · (x − ζpp−1 ) = 1 + x + · · · + xp−1 ,
and letting x = 1, we obtain (1 − ζp )(1 − ζp2 ) · · · (1 − ζpp−1 ) = p. Hence

N(g/m) = N (g)N (m)−1 = pm/p · m−φ(m) = N (R∨ ), (5.61)

39
 ⟨ ⟩
i.e., R∨ = mg .
To prove p/g ∈ R, note that

(1 − ζp )((p − 1) + (p − 2)ζp + · · · + 2ζpp−3 + ζpp−2 ) (5.62)

= (p − 1) − (ζp + ζp2 + ··· + ζpp−1 ) = p, (5.63)

p/g = p/(1 − ζp ) ∈ R. (5.64)

To show that ⟨g⟩ and ⟨p′ ⟩ are coprime for every prime integer p′ ̸= p, note that N (⟨g⟩) =
pm/p , power of p. Since the norm of ⟨g⟩ + ⟨p′ ⟩ is a divisor of both a power of p and of p′ ,
it must be 1, implying that ⟨g⟩ and ⟨p′ ⟩ are coprime.

Remark 5.4.7.
    
 R   R   ⟨g⟩ 
    
 ⟨g⟩ + ⟨p′ ⟩  =  ⟨g⟩   ⟨g⟩ + ⟨p′ ⟩  (5.65)
  
 R   ⟨g⟩ + ⟨p′ ⟩ −1
=     (5.66)
⟨g⟩ ⟨g⟩ 
 
R
Hence, it is a factor of  ⟨g⟩  = pm/p , i.e., a power of p.
On the other hand,
    
 R   R   ⟨p′ ⟩ 
    
 ⟨g⟩ + ⟨p′ ⟩  =  ⟨p′ ⟩   ⟨g⟩ + ⟨p′ ⟩  (5.67)
  
 R   ⟨g⟩ + ⟨p′ ⟩ −1
=  ′    , (5.68)
⟨p ⟩ ⟨p′ ⟩ 
 
 
so it is a factor of  ⟨pR′ ⟩  = p′φ(m) .
∏
Definition
∏ 8. If m = l ml is a product of powers of distinct primes,
⊗define g =
p (1 − ζp ), where p is an odd prime factor of m. For R = Z[ζm ] = l Z[ζml ], let
t = m̂/g ∈ R, where m̂ = m/2 if m even, and m̂ = m otherwise.
∏
Note that m̂/g ∈ R because (1 − ζ2 ) = 2, so m̂/g = m/ p (1 − ζp ) ∈ R, where p runs
over all primes dividing m.

Corollary 5.4.8. R∨ = ⟨g/m̂⟩ = ⟨t−1 ⟩, and ⟨g⟩ is coprime with ⟨p′ ⟩ for every prime
integer p′ except the odd primes dividing m.
⊗
Proof. Just note that R ∼= l Rl , where Rl = Z[ζml ] and g = ⊗gl , where gl = 1 − ζpl .
gl g
R∨ = ⊗l Rl∨ = ⊗l Rl = ⊗l Rl . (5.69)
ml m̂

5.4.5 Prime Splitting and Chinese Remainder Theorem

For an integer prime p ∈ Z, the factorization of principal ideal ⟨p⟩ ⊂ R = Z[ζm ] is as
follows. Let pd be a prime factor of m, let h = φ(pd ), f = the multiplicative order of
p modulo m/pd . Then ⟨p⟩ = ph1 · · · phg , where g = hf
n
, n = φ(m), and pi are distinct

40
primes in R, each of norm pf . (n = φ(m) = φ(pd )φ(m′ ), m′ = m/pd , pf = 1 mod m′ .
Hence, f |φ(m′ )(= n/h). Also, N (⟨p⟩) = pn and N (ph1 · · · phg ) = pf gh = pn .) In particular,
if prime q = 1 mod m, so that q is larger than m, then h = 1 and f = 1, hence ⟨q⟩
splits completely into n distinct prime ideals of norm q in R. Notice that the field Zq
has a primitive mth root of unity, ωm , because the multiplicative subgroup of Zq is cyclic
with order q − 1, which is a multiple of m. Note that ωm i
∈ Zq , where i ∈ Z∗m , are also
distinct mth roots of unity. Then the prime ideal factors of ⟨q⟩ are qi = ⟨q⟩ + ⟨ζm − ωm i
⟩.
Hence, each quotient ring R/qi is isomorphic to Zq via the map ζm 7→ ωm , which confirms
i

N (qi ) = q. In this case,

Z[ζm ] Zq [x] ⊕ Zq [x]
= Zq [ζm ] = = ≈ (Zq )n .
⟨q⟩ Φm (x) i∈Z∗ x − ωm
i
m

∏ Zq [x]
(Note that Φm (x) = i∈Z∗m (x − ωm
i i
), where ωm ∈ Zq , and x−ωm i ≈ Zq for each i ∈ Z∗m ,
because Zq + ωm Zq = Zq .)
i

5.5 Ring-LWE
The formal definition of the ring-LWE problem is provided and the worst-case hardness
result in [LPR10] is shown as follows.
Definition 9 (Ring-LWE Distribution). For a secret s ∈ Rq∨ (or R∨ ) and a distribution
ψ over KR = K ⊗ R, which is isomorphic to H via σ, a sample from the ring-LWE
distribution, As,ψ , over Rq × (KR /qR∨ ) is generated by choosing a ← Rq uniformly at
random, choosing e ← ψ and outputting (a, b = a · s + e mod qR∨ ).
Definition 10 (Ring-LWE, Average-Case Decision). The average-case decision version
of the ring-LWE problem, denoted R − DLW Eq,ψ , is to distinguish with nonnegligible
advantage between independent samples from As,ψ , where s ← Rq∨ uniformly random, and
the same number of uniformly random and independent samples from Rq × (KR /qR∨ ).
Theorem 5.5.1. Let K be the mth cyclotomic number field having dimension n = φ(m),
and R its ring of integers. Let α = α(n) > √0 and let q = q(n) ≥ 2, q = 1 mod m be
a poly(n)-bounded prime such that αq ≥ ω( log n). (Note that √ f = ω(g) if g = o(f ).)
Then there is a polynomial-time quantum reduction from Õ( n/α)-approximate SVIP
(or SVP) on ideal lattices in K to the problem of solving R − DLW Eq,ψ given only l
samples, where ψ is the Gaussian distribution Dξq for ξ = α · (nl/ log(nl))1/4 .
Lemma 5.5.2 (Discretization). Let p and q be positive coprime integers, and ⌊⌉ a valid
discretization, defined earlier, to cosets of pR∨ . Let w ∈ Rp∨ and (a′ , b′ ) ∈ Rq × KR /qR∨ .
Output (a = pa′ mod qR, b) ∈ Rq × Rq∨ , where b = ⌊pb′ ⌉w+pR∨ mod qR∨ . If (a′ , b′ ) ∈
As,ψ , then (a, b) ∈ As,χ where the error distribution χ is ⌊pψ⌉w+pR∨ . If (a′ , b′ ) is uniformly
random, then so is (a, b).
We show that the following variant of ring-LWE is as hard as the original one, closely
following the technique of [ACPS09].
Lemma 5.5.3 (Normal form of R − LW E). Let p and q be positive coprime integers, ⌊⌉
a valid discretization to cosets of pR∨ , and w ∈ Rp∨ . If R − LW Eq,ψ is hard given some
number l of samples, then so is the variant of R − LW Eq,ψ in which the secret is sampled
from χ := ⌊pψ⌉w+pR∨ , given l − 1 samples.

41
Proof. Start by drawing one sample and apply discretization to obtain 0th sample (a0 , b0 ).
Let us assume that the 0th sample (a0 , b0 ) ∈ Rq × Rq∨ is such that a0 is invertible i.e.,
a0 ∈ Rq∗ . From l − 1 samples (ai , bi ) ∈ Rq × KR /qR∨ , (i = 1, . . . , l − 1), output

(a′i = −a−1 ′ ′ ∨
0 ai , bi = bi + ai b0 ) ∈ Rq × KR /qR . (5.70)

This is the same kind of reduction we used to obtain the normal form of standard LWE.
If (ai , bi ) is uniform, so is (a′i , b′i ). If (ai , bi ) ∈ As,ψ , then for each i,

b′i = (ai · s + ei ) − a−1

0 ai (a0 · s + e0 ) (5.71)
′
= ai e0 + ei , (5.72)

where e0 is our secret. Once we find e0 , we obtain s from b0 = a0 · s + e0 .

Remark 5.5.4. When R = Z[ζm ], the fraction of invertible elements in Rq is at least

1/ poly(n, log q) (see the [LPR10]).

Note: As in Z, for prime ideal p of R, an elements a ∈ R is invertible modulo pr if

and
 −1 only if a ̸= 0 mod p. Hence, the fraction of noninvertible elements in R/pr is
R
 ⟨p⟩  = 1/N (p).

The proof of Remark 5.5.4 goes as follows.

Let q = pl11 · · · plαα be a prime-power factorization of q. Then

R ⊕ R
= .
⟨q⟩ ⟨plp ⟩
prime p|q

R R
Note that the fraction of noninvertible elements in is equal to that of ⟨p⟩ . Since
⟨plp ⟩
⟨p⟩ = ph1 · · · phg in R, where h = φ(pd ), f the multiplicative order of p modulo m/pd , pd
the largest power of p that divides m, g = n/(hf ), R = Z[ζm ], n = φ(m), and N (pi ) = pf .
Hence,
∏ n ∏ n
(1 − p−fp ) fp φ(p p ) ≥ (1 − p−fp ) φ(pdp ) .
d
(5.73)
prime p|q prime p|q

Since pfp = 1 mod (m/pdp ), pfp ≥ m

pdp
+ 1, so
n m
φ( )
(1 − p−fp ) φ(pdp ) = (1 − p−fp ) pdp (5.74)
m
≥ (1 − p−fp ) pdp (5.75)
≥ e−1 , (5.76)

using (1 − 1+x ) > e−1 when x > 0. Note that the number of primes diving m is less
1 x

than log2 m. (∵ Let m = pl11 · · · plkk , then log2 m = l1 log2 p1 + · · · + lk log2 pk > k since
li ≥ 1, log2 pi > 1.) Hence, the above product restricted to p which divides both m and
( )log m
q is greater than 1e 2 = poly(m) 1
. If the prime p does not divide m, dp = 0. Hence,
∏
in this case, we compute (1 − p−fp )n because φ(pdp ) = 1. Since pfp are distinct for
p|q,p∤m

42
distinct p and pfp ≡ 1 modulo m, it is bounded below by
log2 q ( )n ( )
∏ 1 ∏
log2 q
1
−n/km
≥ e− α
1
1− ≥ e ∵1− (5.77)
k=1
km + 1 k=1
α+1
∏
log2 q
≥ e−1/k (∵ n = φ(m) < m) (5.78)
k=1
log2 q (
)
∏ 1
−1
≥ e 1− (5.79)
k
k=2
( )
−1 1 2 l−1 1
= (e log2 q) . ∵ · ··· = (5.80)
2 3 l l
1
Thus we have shown that the fraction of invertible elements is greater than poly(n,log q)
.

We used the following fact:

( )x ( )x ( )x+1
1+x 1 1
= 1+ ↗e↙ 1+
x x x
Hence, ( )x ( )x
1 x
1− = ↘ e−1 ,
1+x 1+x
therefore 1 − 1
1+x
> e−1/x for all x > 0.

43
44
Chapter 6

Discrete Fourier Transform &

Chinese Remainder Transform

We follow the algebraic framework of [LPR13].

• ωm : a primitive mth root of unity.

• m: prime power.

• DF Tm : Zm × Zm matrix whose (i, j)th entry is ωm

ij
(i, j = 0, 1, · · · , m − 1).

• CRTm : submatrix of DF Tm obtained by restricting to the rows indexed by Z∗m and

columns indexed by [φ(m)].
∏
For any positive integer with prime factorization m = l ml ,
⊗ ⊗
DF Tm := DF Tml , CRTm := CRTml . (6.1)
l l

√
Remark 6.0.5. DFT is unitary up to scaling by n, while CRT not unitary even up to
scaling.

Decomposition of DF Tm when m is a prime power (Fast Fourier Transform

(FFT))
Let m′ = m/p. We reindex columns of the matrix by j ↔ (j0 , j1 ) ∈ [p] × [m′ ] such that
j = m′ j0 + j1 and rows of the matrix by i ↔ (i0 , i1 ) ∈ [p] × [m′ ] such that i = pi1 + i0 .
(Remark: Let m = pk and write n = αk−1 · · · α1 α0 in p-digit representation, i.e., n =
α0 + pα1 + · · · + pk−1 αk−1 . Then for n = αk−1 αk−2 · · · α1 α0 , j0 = αk−1 , j1 = αk−2 · · · α1 α0 ,
i0 = α0 , i1 = αk−1 · · · α1 .) Then we claim

DF Tm = (I[p] ⊗ DF Tm′ ) · Tm · (DF Tp ⊗ I[m′ ] ), (6.2)

i0 i1
where Tm is a “diagonal” matrix having ωm in the ((i0 , i1 ), (i0 , i1 ))th diagonal entry.
Note that diagonal in this new setting is not diagonal in the standard convention. But
Tm is at least unitary. Also, the matrix multiplication
∑m′ −1 ∑p−1 is defined with respect to the new
(j0 ,j1 ) (α,β) (j0 ,j1 )
column-row index system, i.e., (AB)(i0 ,i1 ) = β=0 α=0 A(i0 ,i1 ) B(α,β) .

45
Proof. Let I[p] ⊗ DF Tm′ = A, Tm = B, and DF Tp ⊗ I[m′ ] = C. Then it suffices to show
that
(j ,j ) (i ,j ) (i ,j ) (j ,j )
(DF Tm )(i00,i11) = A(i00 ,i11) B(i00,j11) C(i00,j11) (6.3)
because of the definitions of A, B, C. Just note that
i1 j1 i0 j1 i0 j0 ′
m i0 j0 +i0 j1 +pi1 j1 ′
(pi1 +i0 )(m j0 +j1 )
ωm ′ ωm ωp = ωm = ωm . (6.4)

Similarly, we have

CRTm = (IZ∗p ⊗ DF Tm′ )T̂m (CRTp ⊗ I[m′ ] ). (6.5)

CRTm is the submatrix of DF Tm restricted to the rows Z∗p × [m′ ] and the columns
[φ(p)] × [m′ ], because Z∗m ∼
= Z∗p × [m′ ] and φ(m) = φ(p) · m′ .
• Z∗p × [m′ ] ↔ i = pi1 + i0 ∈ Z∗m since i0 = 1, · · · , p − 1, and i ∈ Z∗m if and only if i is
not a multiple of p.

• [φ(p)] × [m′ ] ↔ {0, . . . , (p − 2)m′ + (m′ − 1) = pm′ − m′ − 1 = φ(m) − 1}

((j0 , j1 ) ↔ j = m′ j0 + j1 , j0 = 0, · · · , p − 2 = φ(p) − 1)

46
Chapter 7

Powerful basis

7.1 Powerful basis p⃗ of K = Q(ζm) and R = Z[ζm]

• For a prime power m, p⃗T = (ζm
j
)j∈[φ(m)] , a vector over R.
∏
• For m with prime power factorization m = ml , p⃗ = ⊗l p⃗l .

• For I = (R∨ )k ⊂ K of R∨ = ⟨t−1 ⟩, the powerful basis of I is t−t p⃗.

(m/ml )jl
Remark 7.1.1. Note that for p(jl ) = ⊗l ζm
jl
l
, we have, from ζml = ζm ,
∏
p(jl ) ↔ (m/ml )jl
ζm . (7.1)
l

For example, when m = 15, ζ = ζ15 for (j1 , j2 ) ∈ [φ(3) ] × [φ(5) ] ↔ ζ15
5j1 +3j2
, [φ(3)] = {0, 1},
[φ(5) = {0, 1, 2, 3}], the powerful basis consists of

ζ 0 ← (0, 0), ζ 3 ← (0, 1), ζ 5 ← (1, 0), ζ 6 ← (0, 2), (7.2)

ζ 8 ← (1, 1), ζ 9 ← (0, 3), ζ 11 ← (1, 2), ζ 14 ← (1, 3), (7.3)

which are different from the power basis {ζ 0 , ζ 1 , ζ 2 , ζ 3 , ζ 4 , ζ 5 , ζ 6 , ζ 7 }.

Applying the canonical embedding σ, we obtain a Z∗m -by-φ(m) matrix

 
 
 0 1 φ(m)−1 
 σ(ζm ) σ(ζm ) ··· σ(ζm )  (7.4)
 

which is nothing but CRTm , i.e., σ(⃗pT ) = CRTm when m is a prime power.
√ √
Claim: ∥pj ∥∞ = 1 and ∥pj ∥2 = φ(m) = n for all pj .

If m = pk ,
k−1 (p−1)−1
p⃗T = (1, ζm , ζm
2
, · · · , ζm
p
), (7.5)

47
and σ(⃗pT ) is a Z∗m × [φ(m)] matrix such that
 
1 ωm ···
 1 ω2 
 m 
 
σ(⃗pT ) =  ... ...  = CRTm , (7.6)
 
 1 
1
 
ωm
 .. 
 . 
because σ(ζm ) =  i , where i ∈ Z∗m .
 ωm 
..
.

Remark 7.1.2. σ(⃗pT ) is not unitary even up to scaling because σ(pj )s are not orthogonal
to each other, which is same as saying that CRTm is not unitary. Remember that DF Tm
is unitary up to scaling.
T
√
Lemma 7.1.3. The largest √ singular value of σ(⃗
p ) is s1 (⃗
p) = m̂ and the smallest
m
singular value is sn (⃗p) = rad(m) .

Remark
√ 7.1.4.√m̂ = m/2 if m∏is even, otherwise
√ m̂ = m. Note that the ratio of s1 (⃗p) to
φ(m) is just m̂/φ(m) = ( p p−1 ) = O( log log m), where the product runs over
p 1/2

all odd primes dividing m. Note that

∏ ∑ 1 ∑ 1
log2 m ∫ log2 m
p 1
≈1+ ≤1+ ≈1+ dx ≈ log(log m), (7.7)
p−1 p n=1
n 1 x
p|m p|m
prime prime

1
1 √
and that (det R) n = ∆K2φ(m) ≤ φ(m). Hence, p⃗ is a relatively good basis, since s1 (⃗
p)
∥pj ∥
is
√
O( log log m).

Proof. We may assume that m is a power of a prime p. Let m′ = m/p. Then

√
CRTm = ( m′ Q)(CRTp ⊗ I[m′ ] ) (7.8)
√
for some unitary Q, because DF Tm′ / m′ is unitary and so is the Tˆm . Hence, it suffices
to copute the singular values of CRTp . Note that

CRTp∗ CRTp = (pI[φ(p)] − 1 · 1T ). (7.9)

In particular, CRTp is not

( unitary even up )to scaling. To prove this, the first [φ(p)]
1 1 ··· 1
columns of DF Tp is A := . Then
CRTp
 
1
 1 ( )
∗   1 1 ··· 1
A A =  ..  = pI[φ(p)] , (7.10)
 . CRTp∗  CRTp
1

48
 √
because the columns of DF Tp are orthogonal to each other and has length p. Also note
that
( ) ( )
1 1 ··· 1 0 0 ··· 0
A= + . (7.11)
0 CRTp

Then
pI[φ(p)] = A∗ A = CRTp∗ CRTp + 1 · 1T . (7.12)

The eigenvalues and the corresponding eigenvectors of (pI[φ(p)] ) − 1 · 1T are

p ↔ (1, −1, 0, · · · , 0), · · · , p ↔ (1, 0, · · · , −1), (p − 2) times, and 1 ↔ (1, 1, · · · , 1).

7.2 Gram-Schmidt orthogonalization of CRTm

Lemma 7.2.1. Let m be a power of a prime p and m′ = m/p. Then

√
CRTm = Qm ( m′ Dp ⊗ I[m′ ] )(Up ⊗ I[m′ ] ), (7.13)

√
where Qm is unitary, Dp is a real diagonal [φ(p)]-by-[φ(p)] matrix with (p − 1) − j/(p − j)
in its j-th diagonal entry, and Up is an upper unitriangular [φ(p)]-by-[φ(p)] matrix with
−1/(p − i − 1) in its (i, j)th entry 0 ≤ i < j < φ(p).

Proof. We know that

√
CRTm = m′ Q′ (CRTp ⊗ I[m′ ] ) (7.14)

for some unitary Q′ . Thus, it suffices to show that CRTp = Qp Dp Up for some unitary
Qp . We compute
G = CRTp∗ CRTp = (pI[φ(p)] − 1 · 1T ). (7.15)

G has diagonal entries p − 1, and −1 elsewhere. From the uniqueness of Cholesky de-
composition of G, it suffices to show that G = UpT Dp2 Up , where

 . 
.. 0
 √ 
Dp =  p − 1 − j/p − j , (7.16)
..
0 .
 
1 − p−1
1
− p−1
1
· · · − p−1
1

 0 1 − p−2
1
· · · − p−2
1 
 
 
Up =  0 0 1 . (7.17)
 .. 
 . 

49
Let us compute the ith (i ∈ [φ(p)]) diagonal entry in UpT Dp2 Up , which is
∑
(Up )ji (Dp2 )jj (Up )ji (7.18)
j
∑
= (Up )2ji (Dp2 )jj , (7.19)
j
and because of triangularity of Up , we obtain
∑
i−1 ( )
i 1 k
= p−1− + p−1− (7.20)
p − i k=0 (p − k − 1)2 p−k

i ∑
i−1
1
= p−1− +p (7.21)
p−i k=0
(p − k)(p − k − 1)
i
= p−1− + p(T (p) − T (p − i)) (7.22)
p−i
( )
1 1 1 1
where T (k) := + + ··· + =1−
1·2 2·3 (k − 1)k k
i 1 1
= p−1− + p(1 − − 1 − ) (7.23)
p−i p p−i
= p − 1. (7.24)

Computation of the off-diagonal entries is more complicated, but can be done in essentially
the same way.

50
Chapter 8

Chinese Remainder Basis and Fast

Ring Operation

• Note that σ(⃗pT ) = CRTm , hence if a = ⟨⃗p, a⟩, then σ(a) = CRTm a.
R ⊕ R
• Now assume that q is a prime integer = 1 mod m. In this case, = ,
⟨q⟩ i∈Z∗
⟨qi ⟩
m
where qi = ⟨q⟩ + ⟨ζm − ωm
i
⟩ and ωm is some fixed element of order m in Zq .

Definition 11. Chinese remainder (or CRT ) Zq -basis ⃗c of Rq is defined as follows:

• For a prime power m, ⃗c = (ci )i∈Z∗m , where ci = 1 mod qi , ci = 0 mod qj , j ̸= i

(The existence of such ci is guaranteed by the Chinese Remainder Theorem).
∏
• For m having prime-power factorization m = l ml , define ⃗c = ⊗l c⃗l .

For any power I = (R∨ )k of R∨ = ⟨t−1 ⟩, we define t−k⃗c as the CRT Zq -basis of Iq .
Note that the ring operation can be done componentwise if the elements are repre-
sented in the CRT basis, i.e., if a = ⟨⃗c, a⟩ and b = ⟨⃗c, b⟩ ∈ Rq , then the coefficient vector
of a · b with respect to the CRT basis is componentwise multiplication a ⊙ b over Zq by
the defining property of ⃗c. When m is a prime power, the CRT basis ⃗c and the powerful
j
basis p⃗ = (ζm )i∈[φ(m)] are related by

p⃗T = ⃗cT CRTm , (8.1)

j
∑ ij
i.e., ζm = i∈Z∗m ci ωm . To show this identity, just evaluate both sides at qi . Then both
are ωm . They are equal at all of qi , so they are the same. Hence, if a ∈ Rq has the
ij
[φ(m)]
coefficient vector a ∈ Zq in the powerful basis, i.e., a = ⟨⃗p, a⟩, then its coefficient
vector in the CRT basis is CRTm a, i.e., a = ⟨⃗c, CRTm a⟩.

51
52
Chapter 9

Decoding Basis of R∨

−1 m−1
Let τ be an automorphism of R that maps ζm to ζm = ζm . τ is called the conjugation
−1
map since σ(τ (a)) = σ(a). For example, if ζm → e2πi/m
, then ζm 7→ e−2πi/m = e2πi/m .
Note that τ (⃗p) is also a Z-basis of R.

Definition 12. The decoding basis of R∨ is d⃗ = τ (⃗p)∨ , the dual of the conjugate of the
powerful basis p⃗.

Remark 9.0.2. Since R ⊂ R∨ ⊂ KR ⟨and ⟩d⃗ is a basis of R∨ , any a ∈ KR can be

⃗ a for some real vector a. Then
represented in the decoding basis as a = d,

aj = Tr(ad∨j ) = Tr(aτ (pj )) = ⟨σ(a), σ(pj )⟩ ⇐⇒ a = CRTm∗ σ(a), (9.1)

because σ(pj ) is the jth column of CRTm . Since d⃗ is the dual of τ (⃗p), which embeds as
σ(τ (⃗p)) = CRTm , we have σ(d⃗T ) = (CRTm∗ )−1 .

Remark 9.0.3. If L = L(B), then L∨ = L(B−T ).

√
Corollary 9.0.4. The spectral norm of d⃗ is s1 (d)
⃗ = rad(m)/m.

⃗ can be as large as 1, which, unlike p⃗, is much larger than

Remark 9.0.5. s1 (d)

− 1 1
(det R∨ ) n = ∆K 2n ≈ √ ,
1

which may be thought as the average length of a good basis. The decoding basis is
still good choice for discretizing a continuous ring-LWE error, √ because the input error
distribution needs to have Gaussian parameter of at least ω( log n)(≫ 1) for provable
φ(m)−1
worst-case hardness. If d⃗ were defined as the dual of the power basis {1, ζm , · · · , ζm },
then the spectral norm of d⃗ could be much larger: e.g., for m = 1155 = 3 · 5 · 7 · 11,
⃗ ≈ 22.6.
s1 (d)

9.1 Relation to the Powerful Basis

Recall that both d⃗ and t−1 p⃗ are Z-bases of R∨ . We have the following relation between
them.

53
Lemma 9.1.1. Let m be a power of a prime p, and let m′ = m/p, so that φ(m) = φ(p)m′ .
Then
d⃗T = t−1 p⃗T (Lp ⊗ I[m′ ] ), (9.2)
where Lp ∈ Z[φ(p)]×[φ(p)] is the lower triangular matrix with 1s throughout its lower-left
triangle, i.e., its (i, j) entry is 1 for i ≥ j, and 0 otherwise.
Proof. First reindex the conjugate power basis using the index set [φ(p)] × [m′ ], as

τ (p(j0 ,j1 ) ) = ζp−j0 ζm

−j1
.

We have to show that

d(j0 ,j1 ) = t−1 (ζpj0 + ζpj0 +1 + · · · + ζpp−2 )ζm

j1
(9.3)
1 − ζp ζpj0 − ζpp−1 j1
= · · ζm , (9.4)
m 1 − ζp

i.e., the trace of the product of the right hand side with τ (p(j0′ ,j1′ ) ) is 1 if and only if
(j0′ , j1′ ) = (j0 , j1 ). We compute the trace of
1 j0 −j0′ ′ j1 −j1′
(ζp − ζpp−1−j0 )ζm . (9.5)
m
From an earlier computation of Tr(ζm j
), the trace of this is 0 if j1 ̸= j1′ (because j1 −j1′ ̸= 0
mod m′ ), and 0 if j0 ̸= j0′ (because j0 − j0′ , p − 1 − j0′ ̸= 0 mod p. Note that j0 , j0′ =
0, 1, · · · , p − 2.), and otherwise it is m1 (φ(p)m′ − (−m′ )) = m1 (φ(p) + 1)m′ = 1.

9.2 Decoding R∨ and its Powers

Recall the decoding procedure: if Λ is a known fixed lattice and x ∈ H is an unknown
short vector, the goal is to recover x, given t = x mod Λ. Choose {vi } ⊂ Λ∨ a set of
n-linearly independent vectors, not necessarily a basis, which are rather short and let
′
{bi } be a dual basis of {vi }, which generates ∑ a super lattice Λ containing Λ.
′
∑ t mod Λ in the basis {bi } as
Express ci bi , ci ∈ R/Z (so ci = ⟨x, v̄i ⟩ mod 1), then
output Jci Kbi ∈ H. ∑ Then the output equals x if and only if all the coefficients ai = ⟨x, v̄i ⟩
in the expansion x = ai bi are in [−1/2, 1/2). Since (R∨ )∨ = R and every pj of powerful
√
basis of R has ∥τ (pj )∥2 = n, we could use the decoding basis d⃗ for decoding R∨ because
⟨ ⟩
the dual of d⃗ is p⃗ and pj is rather short. But for decoding K/I, where I = (R∨ )k = t−k ,
if we use the Z-basis t1−k d⃗ of I, some elements of the dual of (t1−k⟨d), ⃗ which is tk−1 τ (⃗p),
⟩
∨
might be much longer than the shortest nonzero elements of I = tk−1 . (Remark: Let
t = mg , where m prime, and σ(g) = (1 − ωm 1
, 1 − ωm 2
, . . . , 1 − ωmm−2
). 1 − ωm is very small
1−k ⃗
when m is large. Hence, t is very large.) Instead, we use m̂ d, which generates the
super ideal J = m̂1−k R∨ = t1−k g 1−k R∨ ⊇ I, whose dual elements are m̂k−1 τ (p) ⊂ I ∨ .
Note that √ ∏
∥m̂k−1 τ (⃗p)∥2 m̂k−1 n 1
p−1 )k−1 .
= < ( p (9.6)
λ1 (I ∨ ) λ1 (I ∨ )
odd prime p|m

The last inequality follows from

√ √ (k−1)/n
λ1 (I ∨ ) ≥ nN (R∨ )(1−k)/n = n∆K . (9.7)

54
 Decoding Iq to I, where I = (R∨ )k for some k ≥ 1
⟨ ⟩
For an input ā ∈ Iq , write ā = m̂1−k d, ⃗ ā mod qJ for some ā over Zq , where J =
⟨ ⟩
1−k ∨ 1−k ⃗
m̂ R ⊃ I. Define JāK := m̂ d, JāK if this is in I, otherwise the decoding fails.
⟨ ⟩
Note if a ∈ I, a = m̂1−k d, ⃗ a , and aj ∈ [−q/2, q/2), where aj is jth component of a,
then the decoding succeeds. Hence, if every aj is δ-subgaussian with parameter s, then
by lemma 5.2.1, ⟨Ja mod qIK ⟩ = a except with probability at most 2n exp(δ − πq √/(2s) ).
2 2

Writing a = m̂1−k d, ⃗ a for a ∈ I with integral vector a, we have |aj | ≤ m̂k−1 n∥a∥2 ,
√
because |aj | = |Tr(am̂k−1 τ (pj ))| ≤ ∥a∥2 m̂k−1 n by Schwarz inequality.
If ⟨a is δ-subgaussian
⟩ with parameter s and b ∈ (R∨ )l for some l ≥ 0, we write
ab = m̂1−k−l d,⃗ c for some integral vector c. Then

cj = Tr(m̂k+l−1 τ (pj )ab) (9.8)

= m̂k+l−1 Tr(τ (pj )ba), (9.9)

which is δ-subgaussian with parameter

m̂k+l−1 ∥τ (pj )b∥2 s ≤ m̂k+l−1 ∥τ (pj )∥∞ ∥b∥2 s = m̂k+l−1 ∥b∥2 s. (9.10)

9.2.1 Implementation of Decoding Operation

The goal is to recover an unknown element a ∈ I = (R∨ )k given ā = a mod qI. We
assume ⟩ input ā ∈ Iq is given in the form of a coefficient vector ā over Zq satisfying
⟨ that the
ā = t b, ā mod qI, where ⃗b is some given Zq -basis of R∨ . Output will be given as
1−k⃗
q

a coefficient vector a over Z with respect to the decoding basis t1−k d⃗ of I.

Case 1) k = 1. ⟨ ⟩ ⟨ ⟩
⃗ ā mod qR∨ , output a = d,
If ā = d, ⃗ a where a = JāK.

Case 2) I = (R∨ )k , k > 1.

1. Compute the representation ā′ = ā mod qJ in the Zq -basis m̂1−k⃗b of Jq (recall

that J = m̂1−k R∨ ⊇ I).

2. Decode it as in the case k = 1 to an element a′ ∈ J (which will be equal to a if

successful).

3. Compute the representation of a′ in the Z-basis t1−k d⃗ of I.

For step 1, we want to find ā such that
⟨ ⟩
ā = m̂1−k⃗b, ā mod qJ . (9.11)

k−1 1−k⃗
We claim⟨ that this ⟩ of g ā with respect to the basis t b mod qI,
⟩ ā is the⟨ coefficient
because t1−k⃗b, ā = g k−1 m̂1−k⃗b, ā = g k−1 ā.
1−k ⃗
For
⟨ step 2, ⟩rewrite the output of step 1 with respect⟨ to the basis⟩ m̂ d so that
⃗ ā′ . Then output Jā′ K over Z and let a′ = m̂1−k d,
ā′ = m̂1−k d, ⃗ Ja′ K ∈ J . If it is in I,

55
we succeed. If not, we fail. (Remark: In general, it is easy to decide the membership of
a given lattice.)
For step 3, we convert the representation of a′ in the Z-basis m̂1−k d⃗ of J to a repre-
⃗ Assuming step 2 succeeds, i.e., a′ ∈ I, we want
sentation in a Z-basis of I, namely t1−k d. ⟨ ⟩
to find an integer vector a such that a′ = t1−k d,
⃗ a . For the same a,
⟨ ⟩ ⟨ ⟩
⃗ a = g 1−k a′ ,
⃗ a = g 1−k t1−k d,
m̂1−k d,

i.e., a is the coefficient of g 1−k a′ in the basis m̂1−k d.

⃗

Note that the multiplication by g and the division by g can be computed efficiently.
For example when m = p,

md⃗T = (· · · , (ζpj0 − ζpp−1 ), · · · ), j0 = 0, · · · , p − 2, (9.12)

mg d⃗T = (2 − ζp − ζpp−1 , 1 + ζp − ζp2 − ζpp−1 , · · · ,
1 + ζpp−2 − ζpp−1 − ζpp−1 ), (9.13)
⃗T
md A = (1 − ζpp−1 , ζp − ζpp−1 , · · · , ζpp−2 − ζpp−1 )
 
2 1 1 ··· 1
 −1 1 
 
 −1 1 
×  (9.14)
 .. 
 . 
−1 1
= (2 − ζp − ζpp−1 , · · · ), (9.15)

i.e., g d⃗T = d⃗T A.

9.3 Gaussian sampling in the Decoding Basis

Gaussian sampling a, to be precise σ(a), from KR and representing
⟨ ⟩ it with respect the
decoding basis can be achieved from the fact that if a = d, ⃗ a , then a = CRT ∗ σ(a).
m
∗ ∗
√ ′
Since CRTml = (CRTp ⊗ I[m′l ] ) ml Ql for some unitary Ql ,
l

⊗ √ ⊗
CRTm∗ = (CRTp∗ ⊗ I[m′l ] ) m/ rad(m) Ql . (9.16)
l
l l
∗
Since a spherical Gaussian distribution over H ⊂ CZm is changed into a spherical Gaussian
∗
over H ′ = QH √ ⊂ CZm under the unitary transform Q, it suffices to generate a Gaussian
of parameter s m/ rad(m) over H ′ and then left multiply the result by
⊗
C ∗ := CRTp∗ ⊗ I[m′l ] = CRTrad(m)
∗
⊗ I[m/ rad(m)] . (9.17)
l
l

Since CRTm∗ sends the elements in H to the real vector space of coefficient vectors with
⃗ H ′ ⊂ CZ∗m can be characterized as follows
respect to the decoding basis d,
∗
H ′ = {x ∈ CZm : C ∗ x ∈ R[φ(m)] }. (9.18)

56
For the Gaussian sampling, we have to find a unitary matrix B ′ made up of the elements of
H ′ such that C ∗ B ′ is real. Such B ′ is given in the form Bp′ ⊗Im , since C ∗ = CRTp∗ ⊗I[ml ] .
l l
We show that ( √ )
√−1J
1 I
Bp′ = √ (9.19)
l
2 J − −1I
is one. We check that

(C ∗ B ′ ) = (CRTp∗ Bp′ )ij (9.20)

l l
1
= (e−2πi(ji)/p + e−2πi(j(p−i))/p ) √ (9.21)
2
1 −2πi(ji)/p
= √ (e + e2πi(ji)/p ) ∈ R. (9.22)
2
Note that  
1 1 ··· 1
 
CRTp∗ = 
 −(ij)
.
 (9.23)
ωp

Our B ′ = Bp′ ⊗ Im is different from previous basis of H,

l

( √ )
1 I −1J
√ ∗
B=√ ∈ CZm ×[φ(m)] . (9.24)
2 J − −1I

Even though the Bp′ part looks the same, B ′ is a basis of H ′ , not H.
l

Remark 9.3.1. The final vector of the decoding basis coefficients is C ∗ B ′ c for a real
Gaussian c.

57
58
Chapter 10

Regularity

Let R = Z[ζm ], n = φ(m), and q ≥ 1 a prime. Let a1 , . . . , al−1 be chosen uniformly

and independently from Rq (l could be small). Then we claim that with high probability
∑
over the choice of ai , the distribution of b0 + l−1 i=1 bi ai is within the statistical distance
−Ω(n)
2 of uniform, where bi are chosen from a discrete Gaussian distribution on R of width
essentially nq 1/l . √
nq 1/l is the best possible in some sense, since R is a rotation of nZn , so the discrete
( )n ( )l √
Gaussian of width t covers √tn lattice points. √tn ≈ q n implies t ∼ nq 1/l .
∑
If we consider the more general combination l−1 i=0 bi ai , then the regularity lemma
fails if l is small. For example, when q is a prime satisfying q = 1 mod m, so that ⟨q⟩
splits completely into n ideals of norm q each. Let∑ q denote one of these prime factors.
−l
With probability q all ai are in q, so in this case m i=1 bi ai is in q with certainty whose
distribution is very far from uniformity. By adding the b0 term, we avoid this common
divisor problem.
Lemma 10.0.2. For any n-dimensional lattice Λ and ε, r > 0,
( ( )n )
ηε (Λ∨ )
ρ1/r (Λ) ≤ max 1, (1 + ε). (10.1)
r
Proof. For r ≥ ηε (Λ∨ ), the claim follows from the definition of smoothing parameter
ηε (Λ∨ ). For r < η = ηε (Λ∨ ),

ρ1/r (Λ) = (det Λ)−1 r−n ρr (Λ∨ ) (by Poisson summation formula) (10.2)
( η )n
< (det Λ)−1 r−n ρη (Λ∨ ) = ρ1/η (Λ). (10.3)
r

In particular,
ρ1/r (I) ≤ max(1, N (I)−1 r−n )(1 + 2−2n ), (10.4)
√
since η2−2n (I ∨ ) ≤ n/λ1 (I) ≤ (N (I))−1/n .
Lemma 10.0.3. In the mth cyclotomic number field of degree n, for any q, k ≥ 1,
∑
N (J )k ≤ exp(3c)q kn ≤ q kn+5 , (10.5)
J |⟨q⟩

where c is the number of distinct prime integer divisors of q.

59
Proof. Since c ≤ log2 q and e3c ≤ e3 log2 q < q 5 , the second inequality is trivial. For the
first inequality, we may assume q = pe . Indeed, if q1 and q2 are coprime, then
  
∑ ∑ ∑
N (J )k =  N (J )k   N (J )k  , (10.6)
J |⟨q1 q2 ⟩ J |⟨q1 ⟩ J |⟨q2 ⟩

since when q1 and q2 are coprime, any J | ⟨q1 q2 ⟩ is of the form J = J1 J2 , where J1 | ⟨q1 ⟩,
J2 | ⟨q2 ⟩, and N (J1 J2 ) = N (J1 )N (J2 ), because the ring of integers R is a UFD. Now
⟨p⟩ = ph1 · · · phg in R, where h = φ(pd ), d ≥ 0 is the largest integer such that pd divides
m, each pi is of norm pf , where f ≥ 1 is the multiplicative order of p modulo m/pd , and
g = n/hf , so we have ⟨q⟩ = peh 1 · · · pg , and
eh

∑ ∏
g
N (J ) k
= (1 + N (pi )k + · · · + N (pi )ehk ) (10.7)
J |⟨q⟩ i=1

= (1 + pf k + · · · + pehf k )g (10.8)
≤ pehf kg (1 − p−f k )−g (10.9)
≤ q nk exp(3gp−f k ). (10.10)
Remark 10.0.4.
1 (3x)2
= 1 + x + x2 + · · · < e3x = 1 + 3x + + ··· (10.11)
1−x 2
when x < 12 .
Observe that pf > m/pd , since pf = 1 mod m/pd and pf > 1.
m
g ≤ n/φ(pd ) = φ(m/pd ) < d ,
p
hence gp−f k ≤ gp−f < 1, which proves
∑
N (J )k ≤ q nk e3 . (10.12)
J |⟨q⟩

Hence, for general q, we have

∑
N (J )k ≤ exp(3c)q kn ≤ q kn+5 . (10.13)
J |⟨q⟩

(∵ exp(3c) < exp(3 log2 q) = exp(log2 q 3 ) < q 5 )

[k]×[l]
For a matrix A ∈ Rq , we define
Λ⊥ (A) = {⃗z ∈ R[l] : A⃗z = 0 mod qR}. (10.14)
Theorem 10.0.5. Let R be the ring of integers in the mth cyclotomic number [field K] of
degree n, and q ≥ 2 an integer. For positive integers k ≤ l ≤ poly(n), let A = I[k] |Ā ∈
[k]×[l−k]
(Rq )[k]×[l] , where I[k] ∈ (Rq )[k]×[k] is the identity matrix and Ā ∈ Rq uniformly
random. Then for all r > 2n,
( r )−nl
EĀ [ρ1/r (Λ⊥ (A)∨ )] ≤ 1 + 2 q kn+5 + 2−Ω(n) . (10.15)
n
In particular, if r > 2nq l + nl , then EĀ [ρ1/r (Λ⊥ (A)∨ )] ≤ 1+2−Ω(n) , hence η2−Ω(n) (Λ⊥ (A)) ≤
k 5

r except with probability at most 2−Ω(n) .

60
Corollary 10.0.6. Let R, n, q, k and l as above. Assume A = [I[k] |Ā] ∈ (Rq )[k]×[l] is
chosen as above. Then with probability 1 − 2−Ω(n) over the choice of Ā, the distribution
[k] [l]
of A⃗x ∈ Rq , where each coordinate of ⃗x ∈ Rq is chosen from a discrete Gaussian
distribution of parameter r > 2nq k/l+5/nl over R, satisfies that the probability of each of
the q nk possible outcomes is almost uniform, i.e., is in the interval (1 ± 2−Ω )q −nk .
Proof. Since in this case
η2−Ω(n) (Λ⊥ (A)) ≤ r (10.16)
except with probability at most 2−Ω(n) ,

ρr (Λ⊥ (A) + c) ∈ [1 ± 2−Ω(n) ]rn det(Λ)−1 , (10.17)

[k] [k]
i.e., ∀c ∈ Rq . Hence, every c ∈ Rq occurs almost uniformly, since the probability of
A⃗x = c is proportional to ρ(Λ⊥ (A) + c).
Proof of Theorem. Since x ∈ Λ⊥ (A) ⇔ Ax = 0 mod qR[k] , x ∈ R[l] , and y ∈ Λ⊥ (A)∨ ,
i.e., ⟨y, x⟩ ∈ Z ∀x ∈ Λ⊥ (A), it is easy to see that
{ }
∨ [l] ∨ [k]
(R ) + q A ⃗s : ⃗s ∈ (Rq )
1 T
⊂ Λ⊥ (A)∨ . (10.18)
⟨ ⟩
(For example, 1q AT ⃗s, x = 1q ⟨⃗s, Ax⟩ ∈ Z since Ax ∈ qR[k] .) To show the other inclusion
relation, we consider the simple case, when

A ∈ Zn×m
q , Λ⊥ (A) = {y ∈ Zm : Ay = 0 mod q}. (10.19)

Then { }
⊥ ∨
Λ (A) ⊃ Z + m 1 T
q
A s :s∈ Znq . (10.20)

We assume that A : Zm q → Zq is onto as in our

n
case. Then det(Λ⊥ (A)) = q n , which is
{ }
⊥ ∨
the number of cosets. To show Λ (A) = Z + q A s : s ∈ Zq , it suffices to show that
m 1 T n

the determinant of RHS is 1

qn
. To prove this, assume two translates of Zm ,

Zm + 1q AT s = Zm + 1q AT s′ , s, s′ ∈ Znq . (10.21)

Then 1q AT (s − s′ ) ∈ Zm , so AT (s − s′ ) = 0 mod q. Hence, s − s′ = 0 mod q because

rank(A) = n. That is, the determinant of RHS is q1n because RHS is the union of q n
different translates of Zm . The proof for the general case of Λ is the same.
Now we compute
∑ [ ( )]
EA [ρ1/r (Λ⊥ (A)∨ )] = EA ρ1/r (R∨ )l + 1q AT ⃗s (10.22)
⃗s∈(Rq∨ )[k]
 
∑ ( ) [ ( )]l−k
= ρ1/r (R∨ )[k] + 1q ⃗s  E⃗a ρ1/r R∨ + 1q ⟨⃗a, ⃗s⟩ , (10.23)
⃗s∈(Rq∨ )[k]

where ⃗a represents a typical column vector of Ā, since ∥x∥2 = ∥x1 ∥2 + ∥x2 ∥2 for x =
(x1 , x2 ) ∈ (R∨ )[k] × (R∨ )[l−k] , and

ρ1/r (x) = e−πr∥x∥ = e−πr∥x1 ∥ · e−πr∥x2 ∥ , e−πr∥y∥ = e−πry1 · · · e−πryn .

2 2 2 2 2 2
(10.24)

61
In Eq. (10.22), note that 1q AT ⃗s ̸= 1q AT s⃗′ if ⃗s ̸= s⃗′ in (Rq∨ )[k] , since A is onto, hence AT is
injective.
For any given ⃗s = (s1 , . . . , sk ) ∈ (Rq∨ )[k] , define the ideal

I⃗s = s1 R + · · · + sk R + qR∨ ⊆ R∨ .

Then (Rq )[k] ∋ ⃗a → ⟨⃗a, ⃗s⟩ uniformly random over I⃗s /qR∨ , since ⃗a, which is a column of
⊔ 1 1
Ā, is uniformly random. Since R∨ + ⟨⃗a, ⃗s⟩ = I⃗s ,
∨
q q
⃗a∈I⃗s /qR

∑ ( ) ( )
∨ 1 1
ρ 1 R + ⟨⃗a, ⃗s⟩ = ρ 1 I⃗s , (10.25)
∨
r q r q
⃗a∈I⃗s /qR

where ⃗a is a representative of a different coset element of I⃗s /qR∨ . Hence,

( )
[ ( )] ρ1/r 1q I⃗s
E⃗a ρ1/r R∨ + 1q ⟨⃗a, ⃗s⟩ = . (10.26)
|I⃗s /qR∨ |

Note that when x ∈ {1, · · · , n} is uniformly chosen,

f (x1 ) + · · · + f (xn ) f (x1 ∪ · · · ∪ xn )

Ex f (x) = = , (10.27)
n n
∑
where f (A) = x∈A f (x).
Let T denote the set of all ideals J satisfying qR∨ ⊆ J ⊆ R∨ . Then we can write
Eq. (10.23) as

( )  
∑ ( )l−k ∑ ( )
|J /qR∨ |−(l−k) · ρ1/r 1
q
J · ρ1/r (R∨ )[k] + 1q ⃗s 
J ∈T ⃗s s.t.I⃗s =J
∑ ( )l−k ( ( )k )
∨ l ∨ −(l−k)
≤ ρ 1 (R )(s=0) + |J /qR | · ρ1 1
q
J · ρ 1 q J − 1 . (10.28)
1
r r r
J ∈T \{qR∨ }

Now Eq. (10.28) satisfies

∑ ( )l−k ( ( )k )
∨ l ∨ −(l−k)
ρ 1 (R ) + |J /qR | · ρ1 1
q
J · ρ1 qJ − 1
1
(10.29)
r r r
J ∈T \{qR∨ }
∑ ( ( ) )
l
∨ l ∨ −(l−k)
≤ ρ 1 (R ) + |J /qR | · ρ1 q J − 1
1
(10.30)
r r
J ∈T \{qR∨ }
∑ ( ( ) )
l
∨ −(l−k)
=1+ |J /qR | ρ1 qJ − 1 ,
1
(10.31)
r
J ∈T

so

ρ 1 ( 1q J )l ≤ max(1, (|J /qR∨ |∆K r−n )l ) × (1 + 2−2n )l (10.32)

r

≤ 1 + l21−2n + 2(|J /qR∨ |∆K r−n )l , (10.33)

62
where (10.32) follows from
(( )∨ ) √ ( ( ))−1/n
J n J
η2−2n ≤ (( ) ) ≤ N
q ∨ q
J
λ1 q

and ( )−1    ∨  / 
J  J  R  J 
N =   =   ·  R∨  = ∆K |J /qR∨ |.
q qR R q
Hence,
 
∑  J k
(10.31) < 1 + 2 −Ω(n)
+ 2∆lK r−nl   (10.34)
 qR∨ 
J ∈T
−Ω(n) −nl kn+5
≤ 1+2 + 2(r/n) q , (10.35)

since ∆K ≤ nn and ∑ ∑
|J /qR∨ |k = N (J )k ≤ q kn+5 . (10.36)
J ∈T J |⟨q⟩

J ∨
Note that | qR ∨ | = | qJ ∨ | and qJ
R
⊃ qR.

qR∨ ⊂ J ⊂ R∨ (10.37)
1
q
R ⊃ J∨ ⊃ R (10.38)
R ⊃ qJ ∨ ⊃ qR (10.39)

Conversely,

R ⊃ I ⊃ qR (10.40)
R∨ ⊂ I ∨ ⊂ 1q R∨ (10.41)
qR∨ ⊂ qI ∨ ⊂ R∨ (10.42)

i.e., there is a bijective correspondence

{ideal J : qR∨ ⊂ J ⊂ R∨ } ←→ {ideal I : R ⊃ I ⊃ qR}.

Remark 10.0.7. If qR ⫋ J , then

∑
(R∨ )[k] + 1q ⃗s ⊆ ( 1q J )[k] \ 0, since ⃗s ̸= 0. (10.43)
⃗s s.t.Is̄ =J

Hence,
( ) (( )[k] )
∑ 1 1
ρ 1 (R∨ )k + ⃗s ⊂ ρ1 J −1 (10.44)
q q r q
⃗s s.t.Is̄ =J
( )k
1
= ρ1 J − 1. (10.45)
r q

63
Remark 10.0.8. Another computation:
( ) ∏
k
∨ [k]
ρ 1 (R ) + 1
q
⃗s = ρ 1 (R∨ + 1q si ) (10.46)
r r
i=1
( )
≤ ρ 1 ( 1q J )k−1 · ρ 1 ( 1q J ) − 1 (10.47)
r r

≤ ρ 1 ( 1q J )k − 1, (10.48)
r

where (10.47) follows since si ̸= 0 for some i and R∨ + 1q si ⊂ 1q J , and (10.48) follows
from ρ 1 ( 1q J ) > 1.
r

64
Chapter 11

Cryptosystems

(q should be larger than p, but the smaller the better for the efficiency)

11.1 Dual-Style Cryptosystem [GPV08]

• Gen: choose a0 = −1 ∈ Rq , uniformly random and independent a1 , . . . , al−1 ∈ Rq ,
and independent x0 , . . . , xl−1 ← DR,r . Output
∑
⃗a = (a1 , . . . , al−1 , al = − ai xi ) ∈ Rq{1,2,...,l}
i∈[l]

as the public key, and

⃗x = (x1 , . . . , xl−1 , xl = 1) ∈ R{1,2,...,l}

as the secret key. Note that ⟨⃗a, ⃗x⟩ = x0 ∈ Rq .

• Enc⃗a (µ ∈ Rp ): choose independent e0 , e1 , . . . , el−1 ← ⌊pψ⌉pR∨ and el ← ⌊pψ⌉t−1 µ+pR∨

(shifted error). Let ⃗e = (e1 , . . . , el ) ∈ (R∨ ){1,...,l} . Output the ciphertext

⃗c = e0⃗a + ⃗e ∈ (Rq∨ ){1,...,l} ,

l-samples of Ring-LWE.

• Dec⃗x (⃗c): compute d = J⟨⃗c, ⃗x⟩K ∈ R∨ and output µ = td mod pR.

If r > 2n · q 1/l+2/nl , then (a1 , a2 , . . . , al ) approximating uniform and the above cryp-
tosystem is secure under the hardness R-LWE because ciphertext ⃗c = e0⃗a + ⃗e is a Ring-
LWE with proper security.
Theorem 11.1.1. Suppose that for any c ∈ Rq∨ , ⌊pψ⌉c+pR∨ is δ-subgaussian with pa-
√ √
rameter s for some δ = O( 1l ), and q ≥ s (r2 l + 1)n · ω( log n). Then the decryption
is correct with probability 1 − negl(n) over all the randomness of key generation and
encryption.
Remark 11.1.2. If ψ is continuous Gaussian with parameter
√ s′ > 1 and if we use
⃗ =
coordinate-wise randomized rounding, then since s1 (d) rad(m)
and the sum of two
m
independent Gaussians is again Gaussian with the sum√of variances as the new vari-
ance, ⌊pψ⌉c+pR∨ is 0-subgaussian with parameter s = p s′2 + 2π rad(m)/m = O(ps′ ).

65
(2π rad(m)
m
comes from discretization by coordinate-wise randomized rounding and multi-
plication by d, ⃗ since if E(X) = 0 and |X| ≤ B, then X is 0-subgaussian with parameter
√
B 2π.)
⟨ ⟩
Proof. By construction, ⟨⃗c, ⃗x⟩ = e0 z0 +⟨⃗e, ⃗x⟩ = e , x ⃗′ ⃗′ mod qR∨ , where e⃗′ = (e0 , e1 , . . . , el ),
⟨ ⟩
x⃗′ = (x0 , x1 , . . . , xl = 1), and e⃗′ , x⃗′ = t−1 µ mod pR∨ , so decryption is correct as long
as ⟨ ⟩ ⟨ ⟩
J e⃗′ , x⃗′ mod qR∨ K = e⃗′ , x⃗′ ∈ R∨ . (11.1)
√ √
With high probability, ∥xi ∥2 ≤ r n, ∥xl ∥ = ∥1∥2 = n. Therefore each √ coefficient of
ei xi with respect to decoding basis is δ-subgaussian with parameter sr n, and ⟨ el x⟩l is
√
δ-subgaussian with parameter s n. Hence, each decoding basis coefficient of e⃗′ , x⃗′ is
√
δ(l + 1)-subgaussian with parameters s (r2 l + 1) + n. By decoding Iq to I lemma, this
proves the theorem.

11.2 Compact Public-key Cryptosystem

Let R = Z[ζm ], and p, q coprime integers, Rp the message space. q is coprime with every
odd prime dividing m.
• Gen: choose a uniformly a ← Rq . Choose x ← ⌊ψ⌉R∨ and e ← ⌊pψ⌉pR∨ . Output
(a, b = m̂(ax + e) mod qR) ∈ Rq × Rq as public key, and x as the secret key. (Note
that m̂(ax + e) ∈ gR/gqR, even ax + e ∈ R∨ /qR∨ , because m̂ = tg.)
• Enc(a,b) (µ ∈ Rp ): choose z ← ⌊ψ⌉R∨ , e′ ← ⌊pψ⌉pR∨ and e′′ ← ⌊pψ⌉t−1 µ+pR∨ . Let
u = m̂(za + e′ ) mod qR and v = zb + e′′ ∈ Rq∨ . Output (u, v) ∈ Rq × Rq∨ .
• Decx (u, v): compute v − ux = z m̂(ax + e) + e′′ − m̂(za + e′ )x = m̂(ez − e′ x) + e′′
mod qR∨ , decode it to d = Jv − uxK ∈ R∨ . Output µ = td mod pR.
Theorem 11.2.1. The above cryptosystem is secure assuming the hardness of R −
DLW Eq,ψ .
Proof. If (a, ax + e) ∈ Rq × Rq∨ is indistinguishable from uniform, then (a, m̂(ax + e)) ∈
Rq × Rq is indistinguishable from uniform, since m̂Rq∨ = gR, and ⟨g⟩, ⟨q⟩ are coprime.
Hence, we may assume that the public key (a, b) is uniformly random in Rq × Rq . We
have to prove that (a, b, Enc(a,b) (µ)) is computationally indistinguishable from uniform
for any message µ ∈ Rp . We know that uniform distribution and Az,ψ ( for z ← ⌊ψ⌉R∨ )
over Rq ×KR /qR∨ are computationally indistinguishable from LWE assumption. Now do
the following process. Choose two samples from either uniform or Az,ψ (which we cannot
distinguish). (We may assume that (a, b) is uniform because of R-LWE.) Let them be
(a′ , u′′ ) and (b′ , v ′ ), and apply the discretization process with w = 0 to (a′ , u′′ ) to obtain
(a, u′ ), and with w = t−1 µ ∈ Rp∨ to (b′ , v ′ ) to obtain (b, v). Then output (a, b) as the public
key and (u = m̂u′ mod qR, v) ∈ Rq × Rq∨ as the encryption of µ. If we sampled both
from uniform, then (a, b, u, v) is uniform in Rq × Rq∨ . If we sampled both from Az,ψ , then
[3]

(a, b) is uniform and (u, v) has the same distribution as the one generated by Enc(a,b) (µ).
This means that if we can distinguish random (a, b, u, v) and (a, b, Enca,b (µ)), we can
distinguish uniform distribution and Az,ψ over Rq × KR /qR∨ , which is a contradiction to
the R-LWE assumption. This completes the proof.

66
Lemma 11.2.2. Suppose that ⌊ψ⌉R∨ outputs elements having l2 norm bounded by l
with 1 − negl(n) probability,
√that ⌊pψ⌉e+pR√ ∨ is δ-subgaussian with parameters s for some

δ = O(1), and that q ≥ s 2(m̂l) + nω( log n). Then the decryption is correct with
2

probability 1 − negl(n) over all the randomness of key generation and encryption.

Proof. e, e′ ∈ pR∨ and x, z ∈ R∨ , hence m̂(e·z −e′ ·x) ∈ pR∨ , because m̂ = tg. Therefore
E := m̂(ez −e′ x)+e′′ ∈ R∨ satisfies E = t−1 µ mod pR∨ . So decryption is correct as long
as JE mod qR∨ K = E. By assumption, ∥x∥2 , ∥z∥2 ≤ l with probability 1 − negl(n), and
e, e′ , e′′ are δ-subgaussian with parameter s. Hence, each coefficient of m̂ · ez, m̂ · e′ x ∈ R∨
when represented in the decoding basis√ is δ-subgaussian with parameter sm̂l and √ those of
′′
e are δ-subgaussian with parameter s n (∵ b = 1 in this case, and ∥1∥2 = n). Since
e, e′ , e′′ are mutually
√ independent, each decoding basis coefficient of E is 3δ-subgaussian
with parameter s 2(m̂l)2 + n. The statement follows from decoding Iq to I lemma.

11.3 Homomorphic Cryptosystem

• Notations
R: mth cyclotomic ring of degree n = φ(m)
p, q coprime
Rp : the message space
q: the Ring-LWE modulus
⟨p⟩, ⟨g⟩ ⊂ R coprime, i.e., p is coprime with all primes dividing m

• Gen: s′ ← ⌊ψ⌉R∨ output s = ts′ as secret key.

• Encs (µ ∈ Rp ): e ← ⌊pψ⌉t−1 µ+pR∨ . Let c0 = −c1 s + e ∈ Rq∨ for uniformly random

c1 ← Rq∨ , and output the ciphertext c(S) = c0 + c1 S, where S is indeterminate.
e(= c(s)) is called the noise even though from e, we obtain the message µ (te = µ
mod pR).

• Decs (c(S)) for c of degree k: compute c(s) ∈ (Rq∨ )k and decode it to e = Jc(s)K ∈
R∨ . Output µ = tk e mod pR.

• Homomorphic product is standard polynomial multiplication c(S)c′ (S).

• Homomorphic sum is defined for ciphertexts c, c′ of equal degree as c(S) + c′ (S).

To homomorphically add two ciphertexts of different degrees, we must first homomor-

phically multiply the one having smaller degree by a fixed public encryption of 1 ∈ Rp
enough times to match the larger degree.
From decoding lemma in R∨ , we obtain the following lemma.

Lemma 11.3.1. If q is large enough, or more precisely if the noise e in a degree

√ k√cipher-
text c is δ-subgaussian with parameter r for some δ = O(1), and q ≥ rm̂ k−1
nω( log n),√
then Decs (c) correctly recovers e with probability 1−negl(n). Moreover if q > 2∥e∥2 m̂k−1 n,
then Decs (c) recovers with certainty.

Lemma 11.3.2. The above cryptosystem is secure assuming the hardness of R−DLW Eq,ψ .

67
Proof. We have access to two distributions over Rq × KR /qR∨ , either LWE distribution
As′ ,ψ where s′ ← ⌊ψ⌉R∨ or the uniform distribution. Draw a sample from (a′ , b′ ) ∈
Rq ×KR /qR∨ from the unknown distribution. Let a = pa′ mod qR and b = ⌊pb′ ⌉t−1 µ+pR∨
to obtain (a, b) ∈ Rq ×Rq∨ . Let c1 = −t−1 a ∈ Rq , c0 = b, and output c(S) = c0 +c1 S. If the
unknown distribution is As′ ,ψ , then c(S) is distributed exactly according to Encs (µ). If the
unknown distribution is the uniform distribution, then (a, b) is uniform and independent
of µ. Hence, c(S) uniform. Therefore, if somebody distinguishes the ciphertext c(S)
and the uniform c(S), then he can solve R − DLW E, which contradicts the hardness
assumption of R − DLW E.

11.3.1 Modulus Reduction and Key Switching

We need large modulus q for the correct Decs (c), but for the efficiency, the smaller q is,
the better. For this purpose, we describe modulus reduction.

Let J be a fractional ideal, something like (R∨ )k , and q, q ′ , p integers with both q
and q ′ coprime to p. Let v ∈ Zp be v = q ′ q −1 mod p. Define a randomized function
FJ : Jq → K as follows. Assume that a good basis of J is given and x ∈ Jq . Then FJ (x)
is a short subgaussian element from the coset (v −q ′ /q)x+pJ . Note that (v −q ′ /q)x+pJ
is well defined because (v − q ′ /q)qJ ⊂ pJ . Also observe that for all x ∈ Jq , we have
(q ′ /q)x + FJ (x) ∈ Jq′ up to zero message, i.e., up to a multiple of p. (∵ If x ∈ qJ then
vx = vqJ ⊆ q ′ J , but note that v is defined up to a multiple of p.) It is trivial to see
that qFJ (x) ∈ pJ .

Modulus Reduction Procedure

c(S) = c0 + c1 S is an input ciphertext with c0 , c1 ∈ Rq∨ . Let f0 ← FR∨ (c0 ), f1 ←
t−1 FR (tc1 ), where we used the bases d⃗ for R∨ , and p⃗ for R. Output is c′ (S) = c′0 + c1 S,
where
q′
c′0 = c0 + f0 mod q ′ R∨ , (11.2)
q
q′
c′1 = c1 + f1 (11.3)
q
( ′ )
−1 q
= t tc1 + FR (tc1 ) mod q ′ R∨ . (11.4)
q
Then
q′
c′0 + c′1 s = (c0 + c1 s) + (f0 + f1 s) (11.5)
q
q′
= e + (f0 + (tf1 )s′ ) mod q ′ R∨ . (11.6)
q
′
We define e′ = q ′ /q·e+(f0 +f1 s). Note that e′ = qq e mod pR∨ , since qf0 +q(tf1 )s′ ∈ pR∨ .
The added error term f = (f0 + f1 s) is 0-subgaussian with parameter
√
p 2π(rad(m)/m + m̂∥t−1 s∥2∞ )1/2 , (11.7)
and √ √ √
∥f ∥2 ≤ p n( rad(m)/m + m̂∥t−1 s∥∞ )

68
always if we use coordinate-wise randomized rounding to a coset of pR∨ (respectively,
pR) using the basis pd⃗ (respectively, p⃗p) because of the definition of∑coordinatewise ran-
domized rounding defined at discretization, and the fact that if f = fi bi is the output,
then |fi | ≤ 1.

Key Switching
c(S): degree-k ciphertext,
I = (R∨ )k , d = k + 1,
⃗s = (s0 , . . . , sk ) ∈ R[d] ,
[d]
⃗c ∈ Iq : coefficient vector of a valid degree-k ciphertext c(S), where decryption c(s) =
⟨⃗c, ⃗s⟩ = e mod qI for some short e ∈ t−k µ + pI.

Think of e ∈ I = (R∨ )k as an element in the super lattice m̂1−k R∨ ⊃ I. Then

⟨ ⟩
m̂k−1 e = tm̂k−1⃗c, t−1⃗s . (11.8)

[d]
Let ⃗y = tm̂k−1⃗c ∈ Rq , l = ⌈log2 q⌉, and define

g = (1, 2, 4, . . . , 2l−1 ) ∈ Z[l]

q , (11.9)
G = I[d] ⊗ g T ∈ Z[d]×[dl]
q . (11.10)

[d]
Find short ⃗x ∈ R[dl] such that G⃗x = ⃗y ∈ Rq . (To find such short ⃗x, we do need a good
basis for Λ⊥ (G), which we have; see lemma 23.) We have
⟨ ⟩ ⟨ ⟩
m̂k−1 e = ⃗y , t−1⃗s = ⃗x, t−1 GT ⃗s mod qR∨ . (11.11)

Hint is a collection of independent degree −1 ciphertext hi (S ′ ) for each i ∈ [dl] given

below. (Note that the original secret was (s0 , s1 , . . . , sk ). What we are going to do can
be thought of as the encryption of each sj in an l-vector.)

hi (s′ ) ← Encs′ (0) + t−1 (GT ⃗s)i mod qR∨ , (11.12)

i.e., we generate degree −1 encryptions of 0 and simply add entries of t−1 GT ⃗s to their
constant terms.
hi (S ′ ) = fi + t−1 (GT ⃗s)i (11.13)

for some short fi ∈ pR∨ . For f⃗ = (fi )i∈[dl] , we define

( )2
∑
dl
F := max
∗
|σi (fj )|2 . (11.14)
i∈Zm
j=1

Claim: If all the entries fj ∈ R∨ are δ-subgaussian with parameter s for some δ = O(1),
then √ √
F ≤ Cs · max( dl, ω( log n)) (11.15)
except with negl(n) probability.

69
Proof.
( )
∑
dl
max
∗
|σi (fj )|2 (11.16)
i∈Zm
j=1
( )
∑
dl ∑
dl
= max
∗
Re(σi (fj ))2 + Im(σi (fj ))2 (11.17)
i∈Zm
j=1 j=1
{ dl }
∑ ∑
dl
≤ 2 max
∗
max Re(σi (fj ))2 , Im(σi (fj ))2 . (11.18)
i∈Zm
j=1 j=1

∑
Then previous estimation on P r( i x2i > r) √ implies the claim, since Re(σi (fj )) and
Im(σi (fj )) are δ-subgaussian with parameter s/ 2.
√ √
Remark 11.3.3. √12 (xi + xm−i ) = 2Re(xi ) and similarly for 2Im(xi ), and B =
( √ )
I −1J √ √
√1 √ is unitary basis of H, so 2Re(σ(·)) and 2Im(σ(·)) are Gaussian
2 J − −1I
with parameter s.

Key Switching Procedure

[d] [d]
Input ⃗c ∈ Iq , compute ⃗y = tm̂k−1⃗c ∈ Rq , generate a short ⃗x ∈ R[dl] such that G⃗x = ⃗y .
Output the degree −1 cyphertext
∑
c′ (S ′ ) = xi hi (S ′ ). (11.19)
i∈[dl]

Then
∑
c′ (s′ ) = xi (fi + t−1 (GT ⃗s)i ) (11.20)
⟨ ⟩ ⟨ ⟩
= ⃗x, f⃗ + ⃗x, t−1 GT ⃗s (11.21)
⟨ ⟩
= ⃗x, f⃗ + m̂k−1 e mod qR∨ . (11.22)
⟨ ⟩
Hence, the noise term is e′ = ⃗x, f⃗ + m̂k−1 e. Note that e′ = m̂k−1 e modulo pR∨ , since
fi ∈ pR∨ . e′ is a relatively short element of R∨ , since e was short in m̂1−k R∨ , and ⃗x and
f⃗ are also short in R∨ by construction. To choose a short ⃗x such that G⃗x = ⃗y for a given
⃗y ∈ Rq , it suffices to find a short basis of Λ⊥ (G).
[d]

Lemma 11.3.4. There is an efficiently computable Z-basis Z ∈ R[dl]×[dln] of Λ⊥ (G)

satisfying the following bounds, where ∥Z∥e 2 denotes the largest l2 -norm of the Gram-
√
Schmidt orthogonalized vector e
Z. If q is a power of 2, then s (Z) ≤ 3 e 2=
m̂ and ∥Z∥
√ √ √ 1
2 n, otherwise s1 (Z) ≤ (9 + wt2 (q))m̂ and ∥Z∥ e 2 = 5n, where wt2 (q) denotes the
number of 1s in binary expansion of q.

Proof. Consider the integral lattice

L⊥ (G) = {⃗z ∈ Z[dl] : G⃗z = 0 ∈ Z[d]

q }. (11.23)

70
Define Sg ∈ Z[l]×[l] as  
2
 −1 2 
 
 . 
Sg =  −1 . .  (11.24)
 
 2 
−1 2
if q = 2l , and otherwise
 
2 q0
 −1 2 q1 
 
 .. .. .. 
Sg =  . . . , (11.25)
 
 . .. 2 ql−2 
−1 ql−1
∑
where q = i∈[l] qi 2i is the binary representation of q, with qi ∈ {0, 1}. The columns
of Sq form a basis of L⊥ (g T ), since the columns of Sq are linearly independent and
det Sg = 2l = det(L⊥ (g T )) if q = 2l , and also det Sg = q in general if we consider the
expansion of det Sg with respect to the last column.

The Gram-Schmidt Orthogonalization from earlier with q = 2l is

 
2
 2 
 
Se =  ...  if q = 2l . (11.26)
 
2

If q not a power of 2, we use the standard Gram-Schmidt Orthogonalization. Then

4i 4 − 4−i
∥Sei ∥2 = 1 + ∑ = (< 5) for i = 1, . . . , l − 1, (11.27)
j<i 4j 1 − 4−i
3q 2
∥Sel ∥2 = < 3. (11.28)
4l − 1

(To compute ∥Sel ∥2 , note that Si ⊥ α for i = 1, 2, . . . , l − 1, where α = (1, 2, · · · , 2l−1 ),

hence
⟨sl , α⟩2 q2 3q 2
∥Sel ∥2 = = ∑l−1 j = .) (11.29)
∥α∥2 j=0 4 4l −1

By definition,
∥Au∥
s1 (A) = max . (11.30)
u̸=0 u
When q = 2l ,
   
2 0
 2   −1 0 
   
Sg =  ...  (= A1 ) +  ... ...  (= A2 ).
   
2 −1 0

71
 Sg (u) = A1 u + A2 u, where ∥u∥ = 1. Hence,
∥Sg u∥ ≤ ∥A1 u∥ + ∥A2 u∥ ≤ 2 + 1 = 3.
When q ̸= 2l , we consider SgT .
  
2 −1 u0
 2 −1  u1 
  
 ..  .. 
 .  .  (11.31)
  
 2 −1   ul−2 
q0 q1 ··· ql−2 ql−1 ul−1
 
2u0 − u1
 2u1 − u2 
 
 .. 
=  .  (11.32)
 
 2ul−2 − ul−1 
q0 u0 + · · · + ql−1 ul−1
     
u0 u1 0
 u1   u2   0 
     
 ..   ..   .. 
= 2 .  −  .  +  . , (11.33)
     
 ul−2   ul−1   0 
0 0 q0 u0 + · · · + ql−1 ul−1

where h = (q0 , q1 , · · · , ql−1 ), hence ∥h∥ = wt2 (q). Then

∥SgT u∥2 ≤ (3∥u∥)2 + ∥h∥2 ∥u∥2 ,
so √
∥SgT u∥ < 9 + wt2 (q)∥u∥.
Now we claim that
Z = S ⊗ p⃗T = I[d] ⊗ Sg ⊗ p⃗T ∈ R[dl]×[dln]
is a Z-basis of Λ⊥ (G) satisfying the bounds in the lemma, where p⃗ is the powerful basis
of R. Because
√
s1 (Z) = s1 (S) · s1 (⃗p) = s1 (S) m̂, (11.34)
√
e 2 = ∥S∥
∥Z∥ e 2 ∥CRT
^m ∥2 = ∥S∥ e 2 n, (11.35)
we have proved the lemma.
be given. Then for any Z-basis B of L⊥ (A) ⊂ Z[k] and
[h]×[k]
Remark 11.3.5. Let A ∈ Zq
a Z-basis ⃗b of R, B ⊗ ⃗bT is a Z-basis of Λ⊥ (A) ⊂ R[k] . To prove this, let ⃗z ∈ Λ⊥ (A), i.e.,
[h]
A⃗z = 0 ∈ Rq . Since ⃗z ∈ Rk and ⃗z = (ζ1 , . . . , ζk ), where ζi ∈ R,
∑
ζi = aij bj , i = 1, . . . , k, aij ∈ Z. (11.36)

Then  
a1j
∑  a2j  ∑
 
⃗z = bj  .. = b j · aj ,
 . 
akj

72
  
a1j
 a2j 
 
 = aj . Hence, A⃗z = 0 implies Aaj = 0 ∈ Zq , i.e., aj ∈ L⊥ (A), so aj can
[h]
where  ..
 . 
akj
be written uniquely as a Z-linear combination of basis elements in B, i.e., B ⊗ ⃗bT forms
a basis of Λ⊥ (A) ⊆ R[k] .

73
74
 Part III

Multilinear map

75
Chapter 12

Multilinear maps

12.1 Why multilinear map?

Two-party non-interactive key exchange (2-party NIKE, Diffie-Hellman key exchange)
protocol

• Publish a cyclic group G (i.e., generator g of order q) where discrete log problem is
hard.

• Alice chooses a random x1 ∈ Zq , publishes y1 = g x1

• Bob chooses a random x2 ∈ Zq , publishes y2 = g x2 .

• Alice and Bob compute agreed secret key K = g x1 x2 = y1 x2 = y2x1 .

• Security: Computational Diffie-Hellman problem (CDH), i.e., given g, g x1 , g x2 , com-

pute g x1 x2 .

Wish to have an N -multiparty version: G, GT are groups where Discrete log is hard,
and there is an efficient (N − 1)-linear map e : GN −1 → GT such that

e(g x1 , . . . , g xN −1 ) = e(g, . . . , g)x1 ···xN −1

for all x1 , . . . , xN −1 ∈ Zq .
Then we obtain N -party NIKE:

• Publish cyclic groups G and GT (with generators g and gT , of order q), where
DL-problem is hard, and an efficient (N − 1) linear map e.

• For i = 1, . . . , N , party Pi chooses xi ∈ Zq and publishes yi = g xi .

• All parties can compute the agreed secret key

K = e(y2 , y3 , . . . , yN )x1 (12.1)

= e(g, g, . . . , g)x1 x2 ···xN . (12.2)

• Security: Hardness of Multilinear CDH problem (MCDH), i.e., given g, g x1 , . . . , g xN ,

compute e(g, . . . , g)x1 ···xN .

77
12.2 Grag-Gentry-Halevi (GGH) Graded Encoding
Scheme
High level description

• R = Z[x]/ ⟨xn + 1⟩, where n is a power of 2.

• Publish rings Rg , Rq , and some public parameters of N − 1 Graded Encoding

Scheme.

• For i = 1, . . . , N , party Pi chooses xi ∈ Rg , publishes yi = Enc1 (par, xi : ρi ), i.e.,

level 1 encoding of xi with noise ρi with some encoding scheme.

• We require

Enc1 (par, x1 : ρ1 ) · · · Enc1 (par, xk : ρk ) = Enck (par, x1 · · · xk : ρ) (12.3)

and
x · Enck (par, z : ρ) = Enck (par, x · z : ρ). (12.4)

• Noise-clearing Extraction process at level k

Ext(par, Enck (par, x : ρ)) = r(x) ∈ {0, 1}n (12.5)

should be independent of randomness ρ, and we require output x(x) ∈ {0, 1}n to

be uniform for uniform input x ← U (Rg ).

• Then all parties have agreed secret key

K = Ext(par, EncN −1 (par, x1 · · · xN : ρ)) (12.6)

= Ext(par, x1 y2 · · · yN ). (12.7)

(∵ x1 y2 · · · yN = x1 Enc1 (par, x1 : ρ1 ) · · · Enc1 (par, xk : ρk ) (12.8)

= x1 EncN −1 (x2 · · · xN : ρ) (12.9)
= EncN −1 (x1 x2 · · · xk : ρ)) (12.10)

• Security: Extraction of Graded computational Diffie-Hellman problem (Ext-GCDH):

Given
y1 = Enc1 (par, x1 : ρ1 ), · · · , yN = Enc1 (par, xN : ρN ),
compute
Ext(par, EncN −1 (par, x1 · · · xN : ρ)).

Construction of Enc1 , · · · , Enck

Public parameters

• Sample a small g ←- DR,σ until ∥g −1 ∥ ≤ lg −1 and I = ⟨g⟩ is prime, where DR,σ is

discrete Gaussian with variance σ. Define encoding domain Rg = R/⟨g⟩.

• Sample z ←- U (Rg ).

78
 • Sample a level 1 encoding of 1, i.e., set y = [a · z −1 ]q with a ←- D1+I,σ′ .

• Sample mr level-1 encodings of 0, i.e., set sj = [bj · z −1 ]q with bj ←- DI,σ′ for all
j ≤ mr .

• Sample h ←- DR,√q and define a zero-testing parameter pzt = [ hg z k ]q ∈ Rq .

• Return par = (n, q, y, {xj }j≤mr ) and pzt .

Remark 12.2.1.

R = Z[x]/⟨xn + 1⟩ ↔ Zn (12.11)
∑
n−1
ai xi ↔ (a0 , · · · , an−1 ) (12.12)
i=0
I(ideal) ⊂ R ↔ sublattice of Zn (12.13)

• poly(n)-ideal lattice SVP is assumed to be still difficult even against quantum com-
puter. But note that Gap-SVP for ideal lattice is trivial.

• R could be imbedded in Cn via the canonical map σ = (σ1 , · · · , σn ).

• But when n is a power of 2, the coefficient

√ embedding and the canonical embedding
are isometric up to the constant n.
If α(z) = a0 + a1 z + · · · + an−1 z n−1 , then

(α(ζ), α(ζ 3 ), · · · , α(ζ 2n−1 )) (12.14)

 
1 1 ··· 1
 ζ ζ 3
··· ζ 2n−1 
 
 2 ζ 2·3 ··· ζ 2(2n−1) 
= (a0 , a1 , · · · , an−1 )  ζ , (12.15)
 .. .. .. 
 . . ··· . 
ζ n−1 ζ (n−1)3 ··· ζ (n−1)(2n−1)

where ζ = e2πi/n and n = 2k . Note that

 
1 1 ··· 1
 ζ ζ3 ··· ζ 2n−1 
 
 ζ2 ζ 2·3 ··· ζ 2(2n−1) 
  (12.16)
 .. .. .. 
 . . ··· . 
ζ n−1
ζ (n−1)3
··· ζ (n−1)(2n−1)
√
is unitary up to n. Hence,
√ √ 2
∥σ(α(z))∥ = n |a0 | + · · · + |an−1 |2 (12.17)
√
= n∥α∥. (12.18)

79
Level-1 encoding Enc1 (par, e)
• Given level-0 e ∈ R: e ←- DR,σ′ , u′ = [ey]q , hence u′ = [c′ /z]q with c′ ∈ e + I (Note
that e = [e]⟨g⟩ + geH for some eH ∈ R, where [e]g is the unique coset representative
∑
in Pg , and Pg = { n−1
i=0 ci x g : ci ∈ [− 2 , 2 )}. Also note that (g, xg, · · · , x
i 1 1 n−1
g) is a
short Z-basis of the ideal lattice ⟨g⟩.)

• Rerandomize:
∑mr Sample small ρj ←- DZ,σ1∗ for j ≤ mr , ∑ and return u = [u′ +
′
j=1 ρj xj ]q . Hence, u = [c/z]q with c ∈ e + I and c = c + ρj bj .

Multiplying encodings
Given a level-k1 encoding u1 = [c1 /z1k ]q of e1 and a level-k2 encoding u2 = [c2 /z k2 ]q of e2 ,
u = [u1 · u2 ]q is a level-(k1 + k2 ) encoding of [c1 · c2 ]g . Note that u1 · u2 = [c1 c2 /z k1 +k2 ]q
and c1 · c2 ∈ e1 · e2 + I.

Extraction at level k Ext(par, u)

Given a level-k encoding u = [c/z k ]q , return

1
v = up to lth most significant bit of [pzt u]q with l < ( − ε) log q
4
=: M SBl ([pzt · u]q ).

Correctness of extraction
• At level 1: if c = [c]g + gr for some small r ∈ R, then
( ) ( )
h h
v = M SBl ([c]g + gr) = M SBl [c]g + hr ,
g g
√
which is equal to M SBl ( hg [c]g ) with high probability if q > ∥r∥8 . Since h ∼ q,
√ √ 1 1 5
∥hr∥ ∼ q∥r∥. If q > ∥r∥8 , then q∥r∥ < q 2 + 8 = q 8 . Hence, the noise term
5
hr does not contribute to the most significant lth bit if l < ( 41 − ε) log q, since q 8
contributes up to 58 log q least significant bits.

• After k multiplications: for ui = [ xi +gr

z
i
]q , where i = 1, · · · , k, we have
[ ]
x + gr
u = u1 u2 · · · uk = ,
zk q

where x = x1 · · · xk . Then we require r to satisfy

∥r∥ = O(2k ∥(gr1 ) · · · (grk )∥) (12.19)

(∵ 2k because there are 2k terms)
= O((poly(n)N )k ) < q 1/8 , (12.20)

where N := maxi ∥gri ∥, i.e., O(maxi ∥gri ∥) < q 1/8k / poly(n).

80
Security of GDH for GGH scheme
Known attacks need a small multiple of g, dg (∥dg∥ < q).
Note: From public parameters, it is easy to compute a basis for the ideal ⟨g⟩, even
though g is a secret. But usually the bases thus found are rather long, so it is difficult to
find a short element dg in ⟨g⟩.
[ x+rg ]
Attack on Graded Discrete Log problem. Given u = Enc1 (par, x) = z q
for small r.

• Compute p′zt := [dgpzt ]q = [dg hg z k ]q = [dhz k ]q .

′ 1+ry′ g
• Let u′ = [uy k−1 ]q = [ x+r g
zk q
] , y ′ = [y k ]q = [ zk
]q .

• Compute u′′ := [u′ p′zt ]q = dh(x + r′ · g) ∈ R, y ′′ := [y ′ p′zt ]q = dh(1 + ry′ · g) ∈ R.

• Using a basis for ⟨g⟩ obtained from public parameters, it is easy to compute a (in
general very large) representation x′ ∈ R, where x′ = u′′ y ′′−1 mod ⟨g⟩, so x′ = x
mod ⟨g⟩ since u′′ y ′′−1 = x mod ⟨g⟩.

• Compute a small representation x′′ = x′ mod ⟨dg⟩. Then x′′ = x mod ⟨g⟩.

Note: ⟨dg⟩ is a sublattice of ⟨g⟩, and we have a short basis for the ideal lattice ⟨dg⟩, but
in general not for the ideal lattice ⟨g⟩.

81
82
Chapter 13

GGHLite scheme for k-graded

encoding

Public Parameter Generation

• Sample g ←- DR,σ until ∥g −1 ∥ ≤ lg−1 and I = ⟨g⟩ is prime.

• Sample z ∈ U (Rq ).

• Sample a level-1 encoding of 1: y = [az −1 ]q with a ←- D1+I,σ′ .

• Sample B = (b1 , b2 ) ∈ R × R from (DI,σ′ )2 . If ⟨b1 , b2 ⟩ ̸= I or σn (rot B) < lb , then

resample.
Note:

rot B : R × R → R (13.1)
(x, y) → xb1 + yb2 (13.2)

and σn (rot B) is the smallest singular value of rot B as a linear map.

• Define level-1 encodings of 0:

x1 = [b1 · z −1 ]q , x2 = [b2 · z −1 ]q . (13.3)

• Sample h ←- DR,√q , and define a zero test parameter pzt = [ hg z k ]q ∈ Rq .

• Return parameters = (n, q, y, x1 , x2 , pzt ).

Level-1 encoding Enc(par, e)

Given a level-0 e ∈ R,

• Encode e at level-1: compute u′ = [ey]q .

• Return u = [(u′ + ρ1 x1 + ρ2 x2 )/z]q with ρ1 , ρ2 ←- DR,σ1∗ .

• Hence, u′ + ρ1 x1 + ρ2 x2 is in DI+ρ , σ1∗ B T , ey.

83
Formalizing Re-randomization Security
Informal requirement: Prevent correlation of statistical properties of re-randomized
encoding with encoded element.
Formal requirement: Breaking Ext-GCDH problem is as hard as breaking canonical
Ext-GCDH problem.

• Ext-GCDH: Given public parameters and y1 = [e1 y + ρ11 x1 + ρ21 x2 ]q , · · · , yN =

[eN y + ρ1N x1 + ρ2N x2 ]q , compute

Ext(par, EncN −1 (par, e1 · · · eN : ρ)) = M SBl (pzt · y1 · · · yN ). (13.4)

• Canonical Ext-GCDH: Given public parameters and y1 = [c1 z −1 ]q , · · · , yN = [cN z −1 ]q

with ci ←- DI+ei ,σ∗ B T for i = 1, · · · , N , compute

Ext(par, EncN −1 (par, e1 · · · eN : ρ)) = M SBl (pzt · y1 · · · yN ). (13.5)

Remark 13.0.2. The difference between Ext-GCDH and canonical Ext-GCDH is that
sampling in Ext-GCDH is from a shifted Gaussian (shifted by ei · y), while sampling in
canonical Ext-GCDH is from a fixed origin centered Gaussian, but with a shifted lattice
(by ei ).

Theorem 13.0.3. This requirement is satisfied under suitable parameter conditions.

D1 : The distribution of yi = [vi /z]q in Ext-GCDH problem

• vi distribution is a shifted Gaussian DI+ei ,σ1∗ B T ,c′i with small shifted center c′i = ei y.

D2 : The distribution of yi = [vi /z]q in canonical Ext-GCDH problem

• vi distribution is DI+ei ,σ1∗ B T – which has no shift of center.

The original strong GCDH requirement was based on the statistical distance (SD) ∆:
They required
∑
∆(D1 , D2 ) := |D1 (x) − D2 (x)| < 2−λ . (13.6)
x

Problem with the strong requirement

We ask any adversary A with success probability ε against Ext-GCDH problem. Then
the success probability ε′ is still small (exponentially) against the canonical Ext-GCDH.
Since
|ε − ε′ | < ∆(D1 , D2 ), (13.7)
we have
ε − ∆(D1 , D2 ) < ε′ < ε + ∆(D1 , D2 ). (13.8)
Hence, we need the statistical distance ∆(D1 , D2 ) < 2−λ exponentially small. Conse-
σ∗
quently, we need ∥c′1∥ = 2Ω(λ) (called exponential drowning). Note that
1

84
 F1, Gaussian F2, Gaussian
shifted by c
σ

(c)
∆(F1 , F2 ) = O . (13.9)
σ
Hint:

(Approximated version of the above)

Security analysis of GGHLite is based on Renyi divergence (RD) R

∑
R(D1 ∥D2 ) := D12 (x)/D2 (x) (13.10)
x

Remark on Renyi divergence: On Rn ,

∫ ( )
P 2 (x) P
R(P ∥Q) = dx = EP .
Rn Q(x) Q

Note that
(∫ )2 ∫
P (x)dx P 2 (x)
∫ A
< dx < R(P ∥Q). (13.11)
A
Q(x)dx A Q(x)

(For a general subset√A, the first inequality follows2 from the Cauchy-Schwarz inequality
since P (x) = √P (x)
Q(x).) Hence, Q(A) ≥ P (A) /R(P ∥Q).
Q(x)

85
Security analysis of GGHLite based on Renyi divergence
Any adversary A with success probability ε against Ext-GCDH problem has success
probability ε′ against canonical Ext-GCDH problem with
ε′ ≥ ε2 /R(D1 ∥D2 )2 . (13.12)
Hence, we require only that R(D1 ∥D2 ) is poly(λ). Then ϵ′ ∼ 2−λ implies ε ∼ 2−λ .
Lemma 13.0.4. For any n-dimensional lattice Λ ⊂ Rn and rank n matrix S ∈ Rm×n , let
P be the center-shifted Gaussian distribution DΛ,S,w , and Q the center-shifted Gaussian
distribution DΛ,S,z for some w, z ∈ Rn . If w, z ∈ Λ, let ε = 0. Otherwise fix ε ∈ (0, 1)
and assume that σn (S) ≥ ηε (Λ). Then
[( )2 ( )2 ]
1−ε 1+ε
R(P ∥Q) ∈ , exp(2π∥S −T (w − z)∥2 ) (13.13)
1+ε 1−ε
[( )2 ( )2 ] ( )
1−ε 1+ε 2π∥(w − z)∥2
⊂ , exp (13.14)
1+ε 1−ε σn (S)2
(refer the paper [LSS14] on GGHLite for the proof.)
Hence, the lemma implies that
( )
2π∥c′1 ∥2
R(D1 ∥D2 ) ≤ exp . (13.15)
σn (σ1∗ B T )2
σ1∗
For the requirement R(D1 ∥D2 ) ≤ poly(λ), we can use c′1
= O(1/| log λ|).

In our scheme, vi = [ei a + ρ1 b1 + ρ2 b2 /z]q with ρi ←- DR,σ1∗ . Hence, we have to show

that
ρ1 b1 + ρ2 b2 ≈ DI,σ1∗ B T , (13.16)
where B = [b1 , b2 ] = g[t1 , t2 ] ∈ R2 .
• Step 1. We show [t1 , t2 ]R2 = R with nonzero probability.
Remark 13.0.5. It is the probability that two random algebraic integers are co-
prime ≈ ζR (2)−1 as in the integer case. (Let n1 , n2 be random integers, and p a
prime. Then the probability that both n1 and n2 has p as a common factor is p12 .
∏ ( )
Hence, the probability that the pair n1 , n2 is coprime is given by p prime 1 − p2 ,
1

which is ζ(2)−1 .)
Remark 13.0.6. [t1 , t2 ]R2 ̸= R is non-negligible for R = Z[x]/⟨xn + 1⟩, where n
is even, since each random element of R falls in the ideal ⟨x + 1⟩ with probability
1
2
, hence both t1 , t2 get stuck in ⟨x + 1⟩ with probability 14 . (h = a0 + a1 x + · · · +
an−1 xn−1 ∈ R is defined up to a multiple of (xn + 1), i.e., if h̃ = h + f (x)(xn + 1)
in R for some polynomial f (x), then h̃ = h in R. Hence, h ∈ ⟨x + 1⟩ if and only if
h(−1) = 0 for some f (x), i.e., there exists f (x) such that
a0 − a1 + · · · ± an−1 + f (−1)2 = 0, (13.17)
that is, if a0 − a1 + · · · ± an−1 is even.)
• Step 2. Let AT = {V ∈ R2 : T V = [t1 , t2 ]V = 0}. If σ1∗ > ηε (AT ), then ρ1 t1 + ρ2 t2 is
within SD 2ε of DR,σ1∗ T t , which comes from discrete Gaussian leftover hash lemma.

86
Chapter 14

Cryptanalysis of GGH map

We follow the notations in Steinfeld’s lecture slides. We only explain essential parts of
the cryptanalysis due to Yupu Hu and Huiwen Jia [Hu15].

14.1 Schematic description of the cryptanalysis

1. From public noised encoding V of secret v, one generates an equivalent secret, v (0) ,
of which the noise v (0) − v is not short in general and v (0) − v ∈ ⟨g⟩.
∏
K+1 ∏
K+1 ∏
K+1
2. For the product v (k)
, v (0,k)
− v (k) ∈ ⟨g⟩, but not short.
k=1 k=1 k=1

3. Key step of modified encoding/decoding.

∏
K+1 ∏
K+1
From η := v (0,k) = v (k) + ξg, we obtain
k=1 k=1

∏
K+1
′′′ K −1
η = (h(1 + ag) g ) v (k) + ξ ′′ (1 + ag) mod q,
k=1

where ξ ′′ (1 + ag) short. Hence, the higher order bits of η ′′′ are what we want to
obtain.

14.2 Generating an equivalent secret

Y := y K−1 x(1) pzt (mod q) (14.1)
= h(1 + ag)K−1 b(1) (14.2)
X (i) := y K−2 x(i) x(1) pzt (mod q) (14.3)
= h(1 + ag)K−2 (b(i) g)b(1) (14.4)
Note that RHSs are rather short.
V → W := V y K−2 x(1) pzt (mod q) (14.5)
= vY + (u(1) X (1) + u(2) X (2) ) : short (14.6)
→ W (mod Y ) = (u(1) X (1) (mod Y ) +
u(2) X (2) (mod Y )) (mod Y ) (14.7)

87
 ⟨ ⟩
From W (mod Y ), X (1) (mod Y ), and X (2) (mod Y ), obtain W ′ ∈ X (1) , X (2) such that
W − W ′ (mod Y ) = 0.
Denote W ′ = u′(1) X (1) + u′(2) X (2) .
v (0) := (W − W ′ )/Y (14.8)
= v + ((u(1) X (1) + u(2) X (2) ) − W ′ )/Y (14.9)
= v + ((u(1) − u′(1) )X (1) + (u(2) − u′(2) )X (2) )/Y (14.10)
/
= v + ((u(1) − u′(1) )b(1) + (u(2) − u′(2) )b(2) g) (1 + ag). (14.11)
Since g and 1 + ag are coprime,
v (0) − v ∈ ⟨g⟩ .
v (0) is called the equivalent secret of v.

14.3 Modified Encoding/Decoding

∏
K+1 ∏
K+1
(0,k)
η := v = v (k) + ξg (14.12)
k=1 k=1
∏
K+1
′
η := Y η = Y v (k) + ξ ′ b(1) g (14.13)
k=1
′′ ′ (1)
η := η (mod X ) (14.14)
∏
K+1
η ′′ = Y v (k) + ξ ′′ b(1) g (14.15)
k=1

(∵ η ′′ is the sum of η ′ and a multiple of X (1) , and X (1) is a multiple of b(1) g.) Note that
√ (1) ∏
K+1
′′ (1)
η has size nX by the definition (mod X ), and that Y v (k) also small. Hence,
k=1
∏
K+1
ξ ′′ b(1) g = η ′′ − Y v (k) is small.
k=1

η ′′′ := y(x(1) )−1 η ′′ (mod q) (14.16)

∏
K+1
K −1
= (h(1 + ag) g ) v (k) + ξ ′′ (1 + ag) (mod q) (14.17)
k=1

∏
K+1
′′ K −1
Note that ξ (1 + ag) is small and that (h(1 + ag) g ) v (k) (mod q) is the decoded
k=1
message, so its high order bits are what we want to obtain.

14.4 Witness encryption based on 3–exact cover

• 3–exact cover problem:
A subset of {1, 2, . . . , 3K} which has exactly three elements is called a piece. A

88
 collection of K pieces without intersection is called a 3–exact cover of {1, 2, . . . , 3K}.
The 3–exact cover problem is that, for randomly given N (K) different pieces with
a hidden 3–exact cover, find it. If N (K) = O(K), it easy. If N (K) = O(K 2 ), it is
hard.

• Encryption:

– Sample short elements v (1) · · · v (3K) ∈ R.

– Compute v (1) · · · v (3K) y K pzt (mod q).
– Then EKEY is its high-order bits.
Hide EKEY into pieces as follows. Randomly generate N (K) different pieces of
{1, 2, . . . , 3K} with a hidden 3–exact cover called EC. For each piece {i1 , i2 , i3 },
compute noised encoding of v (i1 ) v (i2 ) v (i3 ) , i.e.,

V {i1 ,i2 ,i3 } = v (i1 ) v (i2 ) v (i3 ) y + (u({i1 ,i2 ,i3 },1) x(1)
+u({i1 ,i2 ,i3 },2) x(2) ) (mod q). (14.18)
∏
• Decryption: If one knows EC, compute pzt V (i1 ,i2 ,i3 ) (mod q). Then
{i1 ,i2 ,i3 }∈EC
EKEY is its high-order bits.

14.5 Breaking WE based on the hardness of 3–exact

cover problem
Given N (K) = O(K 2 ) different pieces of {1, 2, . . . , 3K}, {i1 , i2 , i3 } is called a combined
piece if

1. {i1 , i2 , i3 } is not a piece given;

2. {i1 , i2 , i3 } = ({j1 , j2 , j3 } ∪ {k1 , k2 , k3 }) − {l1 , l2 , l3 };

3. {j1 , j2 , j3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } are pieces given.

{i1 , i2 , i3 } is called a second-order combined piece if

1. {i1 , i2 , i3 } is neither a piece nor a combined piece;

2. {i1 , i2 , i3 } = ({j1 , j2 , j3 } ∪ {k1 , k2 , k3 }) − {l1 , l2 , l3 };

3. {j1 , j2 , j3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } are pieces given or combined pieces.

We define combined 3–exact cover of {1, 2, . . . , 3K} as before.

The combined 3–exact cover problem is to “find a combined 3–exact cover among pieces,
combined pieces, and second-order combined pieces.”

Claim: The combined 3–exact cover problem is easy.

We are given random K 2 pieces, and there is a hidden 3–exact cover among them. Then
for a random {i1 , i2 , i3 } which is not piece,

Prob({i1 , i2 , i3 } is not a combined piece) ∝ e−1 ,

89
as shown below. Hence, from the random K 2 pieces, we obtain about (1 − e−1 )C33K
different subsets of {1, 2, . . . , 3K} which are pieces or combined pieces. There are about
e−1 C33K left over 3-element subsets of {1, 2, . . . , 3K} which are neither pieces nor combined
pieces. Choose one {i1 , i2 , i3 } from them. We show that

Prob({i1 , i2 , i3 } is not a second-order piece) ∝ e−K

3

by the same method of computatiing Prob({i1 , i2 , i3 } is not a combined piece). Hence,

almost all of different C33K subsets of {1, 2, . . . , 3K} which consists of 3 elements are
pieces, combined pieces, or second-order combined pieces, so the combined 3–exact cover
problem is easily solved by choosing a random 3–exact cover among pieces, combined
pieces, and 2nd order combined pieces.

Computation of Prob({i1 , i2 , i3 } is not a combined piece)

Take a random {i1 , i2 , i3 } which is not a piece and fix it. Take a random {α, β, γ} from
{1, 2, . . . , 3K} − {i1 , i2 , i3 }. Then partition {α, β, γ, i1 , i2 , i3 } into two parts, {j1 , j2 , j3 }
and {k1 , k2 , k3 }. Let {l1 , l2 , l3 } = {α, β, γ}. The number of possibilities of such 3-triples
{j1 , j2 , j3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } is C33K−3 · C36 . Hence, the probability out of all possible
3-triples is
C33K−3 · C36 80 1
≈ ≈ 6.
3K 3 81K 6 K
(C3 )
Hence, the probability that there is no 3 triples {j1 , j2 , j3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } such
that
{j1 , j2 , j3 } ∪ {k1 , k2 , k3 } ⊃ {i1 , i2 , i3 }
{l1 , l2 , l3 } = {j1 , j2 , j3 } ∪ {k1 , k2 , k3 } − {i1 , i2 , i3 }
among 3 random pieces {j1 , j2 , j3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } is about
( )O(K 2 )(O(K 2 )−1)(O(K 2 )−2) ( )
1 O(K 2 )3
1− 6 ≈ exp − 6
≈ e−1 .
K K
Take a fixed combined 3–exact cover. Take an element {i1 , i2 , i3 } of this combined
3–exact cover.
1. If {i1 , i2 , i3 } is a piece, we count +1.
2. If {i1 , i2 , i3 } is a combined piece, so that
{i1 , i2 , i3 } = {j1 , j2 , j3 } ∪ {k1 , k2 , k3 } − {l1 , l2 , l3 },
count {i1 , i2 , i3 } → +1, {k1 , k2 , k3 } → +1, {l1 , l2 , l3 } → −1.
3. If {i1 , i2 , i3 } is a second-order combined piece, then
{i1 , i2 , i3 } = {j1 , j2 , j3 } ∪ {k1 , k2 , k3 } − {l1 , l2 , l3 },
where {i1 , i2 , i3 }, {k1 , k2 , k3 }, {l1 , l2 , l3 } are pieces given or combined pieces.
(3–1) If {i1 , i2 , i3 } is a piece given, count +1.
If {i1 , i2 , i3 } is a combined piece, count
{i1 , i2 , i3 } = {α1 , α2 , α3 } ∪ {β1 , β2 , β3 } − {γ1 , γ2 , γ3 }.
+1 +1 −1

90
 (3–2) Similarly for {k1 , k2 , k3 }.
(3–3) If {l1 , l2 , l3 } is a piece given, count −1.
If {l1 , l2 , l3 } is a combined piece, count
{l1 , l2 , l3 } = {ϵ1 , ϵ2 , ϵ3 } ∪ {δ1 , δ2 , δ3 } − {ξ1 , ξ2 , ξ3 }.
−1 −1 +1

Note: It is possible that one piece is counted several times.

CP F = collection of all positive factors

N P F = the number of positive factors

CN F and N N F are similarly defined.

Remark 14.5.1. N P F − N N F = K.

Since there are about K 2 pieces with factors (+1), there are (1 − e−2 )C33K − K 2
combined pieces with factors (+, +, −), and e−2 C33K second-order combined pieces with
factors at most (+ + + + +, − − − −). Hence, for a randomly chosen combined 3–exact
cover, it is almost certain that N P F ≤ 3K, hence N N F ≤ 2K.
( )
K2 K2
(∵ 5e + 2 · 1 − e · 3K + 1 · 3K ≤ 2 + 3e−2 < 3)
−2 −2
C3 C3

If all of our combined 3–exact cover consists of pure pieces, N P F − N N F = N P F = K.

If one of pure pieces is replaced by a combined piece, N P F − N N F is not changed. The
same result holds for second-order combined pieces.

Breaking WE
Randomly take a combined 3–exact cover → Obtain CP F and CN F .
For a positive factor (pf ) = {i1 , i2 , i3 }, denote the secret of (pf ) as v (pf ) = v (i1 ) v (i2 ) v (i3 ) ,
and the equivalent secret of v(pf ) as v ′(pf ) .
∏
P P F := v ′(pf ) (14.19)
pf ∈CP F
∏
P N F := v ′(nf ) (14.20)
nf ∈CN F

∏
3K
P T S := v (k) (14.21)
k=1

Then
∏ ∏
1. v (pf ) = P T S × v (nf )
pf ∈CP F nf ∈CN F
∏
2. P P F − v (pf ) ∈ ⟨g⟩
pf ∈CP F
∏
3. P N F − v (nf ) ∈ ⟨g⟩
nf ∈CN F

91
 4. P P F − P N F × P T S ∈ ⟨g⟩
If P T S ′ is an equivalent secret of P T S, then P P F − P N F × P T S ′ ∈ ⟨g⟩, since P T S ′ −
P T S ∈ ⟨g⟩. Conversely, if P N F and g are coprime, and if P P F − P N F × P T S ′ ∈ ⟨g⟩,
then P T S ′ is an equivalent secret of P T S, since in this case P N F × (P T S ′ − P T S) ∈ ⟨g⟩
implies (P T S ′ − P T S) ∈ ⟨g⟩.
Note that the Hermite normal form of
 
g0 g1 · · · gn−1
 −gn−1 g0 · · · gn−2 
 
g= .. .. ... ..  , (14.22)
 . . . 
−g1 −g2 · · · g0

which is the matrix representation of (g, xg, . . . , xn−1 g)T , is

 
G0
 G1 1 
 
G =  .. . , (14.23)
 . . . 
Gn−1 1

where G0 is the absolute value of det g, and Gi (mod G0 ) = Gi . This can be obtained by
Gauss elimination once the basis of ⟨g⟩ is formed. Hence,

P P F − P N F × P T S ′ ∈ ⟨g⟩ (14.24)
⇔ P P F G−1 − P T S ′ × P N F × G−1 ∈ R, (14.25)

where  
P N F0 P N F1 · · · P N Fn−1
 −P N Fn−1 P N F0 · · · P N Fn−2 
 
PNF =  .. .. .. . (14.26)
 . . . 
−P N F1 −P N F2 · · · P N F0
Let lcm be the least common multiple of all denominators of the entries of P P F G−1
and P N F × G−1 . Then

(lcm × P P F × G−1 ) (mod lcm) (14.27)

= P T S ′ × (lcmP N F × G−1 ) (mod lcm). (14.28)

Note that there is at least one solution, namely P T S, which we do not know. Obtain a
solution P T S ′ . Let η = P T S ′ , and compute η ′ = Y η. Let η ′′ = η ′ (mod X (1) ), and again
compute η ′′′ = y(x(1) )−1 η ′′ (mod q). The high-order bits of η ′′′ is then what we wanted.
Remark 14.5.2. We must obtain the Hermite normal form of ⟨g⟩ for an unknown
small
⟨ g, when Y ,⟩ X (1) ,⟨X (2) are public. ⟩First we obtain the Hermite normal
⟨ ⟩forms
⟨ of⟩
h(1 + ag)K−2 b(1) and h(1 + ag)K−2 b(1) g when the principal ideals ⟨Y ⟩, X (1) , X (2)
are known. Note that if the Hermite normal form of the principal ideal ⟨g ′ ⟩ is
 
G′0
 G′ 1 
 1 
 .. . 
 . .. 
′
Gn−1 1

92
and g is a factor g ′ , then the Hermite normal form of ⟨g⟩ is
 
G0
 G′ (mod G0 ) 1 
 1 
 .. .. ,
 . . 
′
Gn−1 (mod G0 ) 1

where G0 is the determinant of ⟨g⟩.

14.6 Computing the Hermite Normal Form of ⟨g⟩

by
⟨ computing ⟩the Hermite
⟨ Normal⟩ Forms of
K−2 (1)
h(1 + ag) b and h(1 + ag)K−2b(1)g
We assume that 1 + ag and b(1) g are coprime.

1. Gaussian sample Z from the lattice ⟨Y ⟩.

2. Compute
⟨ Z ′ = Z mod⟩ X (1) . Then Z ′ is uniformly distributed over the intersection
area h(1 + ag)K−2 b(1) ∩ P P (X (1) ). (Since 1 + ag and b(1) g are coprime, multipli-
cation (or division) by 1 + ag preserves the uniformity over P P (X (1) ).)

3. Compute the determinant of ⟨Z ′ ⟩.

4. Repeat the above steps several times.

5. Compute the greatest common divisor of the polynomially

⟨ many ⟩ sample. Then with
K−2 (1)
a high probability, it is the ⟨determinant of h(1 ⟩ + ag) b . Hence, we obtain
the Hermite
⟨ (1) ⟩normal′ form of h(1 + ag)
K−2 (1)
b from that of ⟨Y ⟩. Z⟨ ′ is of⟩ the form
⟨⟨Y ⟩ − X K−2, so ⟩ is in the greatest common divisor of ⟨Y ⟩ (1)
(1)
Z and X , which is
(1)
h(1 + ag) b , and of course in the parallelepiped P P (X ).
⟨ ⟩
Similarly, we⟨ obtain
⟩ the Hermite normal form of h(1 + ag)K−2 b(1) g by sampling from
the lattice X (2) . Then compute mod X⟨(1) , . . . ⟩ ⟨ ⟩
With the two Hermite normal forms of h(1 + ag)K−2 b(1) and h(1 ⟨+ ag)K−2 b(1) g , we⟩
obtain the Hermite normal⟨ form ofK−2 ⟨g⟩. (Just
⟩ divide the determinant of h(1 + ag)K−2 b(1) g
by the determinant of h(1 + ag) b(1) , then obtain the determinant of ⟨g⟩.)

Remark 14.6.1. Recently a quantum algorithm was found that can compute small
generators of principal ideals in the cyclotomic ring. (In particular, Soliloquy; Campbell,
Groves, Shepherd. [Cam14]) That is, small generators themselves of ⟨g⟩ are found, not
only the secrets of multipartite NIKE or WE. But the cryptanalysis of Hu and Jia are
classical analysis.

93
94
Appendix A

Hermite Normal Form of Ideal

Lattices (following Ding and
Lindner, Smart and Vercauteren)

Let I ⊆ R = Z[x]/⟨f (x)⟩ be an ideal, where f (x) = a0 + a1 x + · · · + an−1 xn−1 + xn

is an irreducible monic polynomial. Let L be the corresponding ideal lattice under the
coefficient embedding, and B ∈ Zn×n a basis of L. Then we have
 
0 0 ··· 0 −a0
 .. 
 .  B = BT
In−1
−an−1

for some integral matrix T , because it corresponds to the invariance of I under the
multiplication by x. If B is the HNF-basis of L, then the diagonal entries form a division
chain
B(n,n) | B(n−1,n−1) | · · · | B(1,1) ,
 
  0
∗  ∗ 
 B(i,i)   
   B(i,i) 
   
because when the ith column  0  is multiplied by x, it becomes  0 , and
 ..   
 .   .. 
 . 
0
0
 
0
 ∗ 
 
 B(i,i) 
 
it should be a linear combination of B over integers, i.e.,  0  = Bt for some
 
 .. 
 . 
0
integral vector t. Comparing both sides, especially the (i + 1)th component, we have
B(i+1,i+1) ti+1 = B(i,i) , showing that B(i+1,i+1) | B(i,i) .
When I = ⟨p, x − α⟩, two element representation of I, where p is the norm of I and α
is a root of f (x) modulo p, the corresponding HNF representation is very simple. Since

p, px, . . . , pxn−1 , (x − α), x(x − α), . . . , x2 (x − α), . . . , xn−1 (x − α)

95
are all in the ideal I and span I, we obtain HNF of the ideal lattice L,
 
p −α −α2 · · · −αn−1
 0 
 
 0 
 ,
 .. In−1 
 . 
0

where all integers in the first row, and in the second column and onward, are taken
modulo p. But it is a bad basis of ideal lattice I, in general.

96
Appendix B

Notes on Cyclotomic Fields with

Examples (by H. Kim)

B.1 Cyclotomic Number Fields & Ring of Integers

The group of units For m = 2k+1 , the group of units in Zm is given by
{ }
Z∗m = 1, 3, 5, 7, 9, . . . , 2k+1 − 1 ,

so n := φ(m) = |Z∗m | = 2k .

Cyclotomic number fields & ring of integers The minimal polynomial over Q of
a primitive mth root of unity is called the mth cyclotomic polynomial, and it is denoted
by Φm (x). Since Φm (x) | xm − 1, the coefficients of Φm (x) are in Z by Gauss’s Lemma,
i.e., Φm (x) ∈ Z[x].
When m = 2k+1 , it is given by
k
Φm (x) = x2 + 1,

so the field extension of Q by an mth root of unity is

k
Q(ζm ) = Q[x]/Φm (x) = Q[x]/(x2 + 1).
2 −1
Q(ζm ) is therefore a degree 2k field extension over Q, and 1, ζm , ζm is a Q-basis.
2 k
, . . . , ζm
An element of Q(ζm ) is said to be integral (over Z) if it is the root of a monic
polynomial with integer coefficients. For example, ζm is integral since it is the root of
the polynomial xm − 1. The ring of integers for Q(ζm ), i.e., the set of integral elements
of Q(ζm ), is given by
k
R := Z[ζm ] = Z[x]/Φm (x) = Z[x]/(x2 + 1).

R is a Dedekind domain, and a free abelian group of rank 2k . Any Z-basis of R is a

Q-basis of Q(ζm ), since linear independence over Z is equivalent to linear independence
2k −1
over Q. Any Z-basis of R is called an integral basis. 1, ζm , ζm
2
, . . . , ζm ∈ R is an integral
basis, called the power basis. Note that Q(ζm ) is a field of fractions for R. Since R is
a Dedekind domain, every nonzero ideal of R can be written uniquely as a product of
prime ideals.

97
Examples

• m=4

– Z∗m = {1, 3}
– n=2
– Q(ζm ) = Q[x]/(x2 + 1) = Q(i)
– R = Z[x]/(x2 + 1) = Z[i]
– Power basis: {1, i}

• m=8

– Z∗m = {1, 3, 5, 7}
– n=4
( )
– Q(ζm ) = Q[x]/(x4 + 1) = Q 1+i √
2
[ ]
– R = Z[x]/(x4 + 1) = Z 1+i√
2
{ }
√ , i, −1+i
– Power basis: 1, 1+i √
2 2

Note that Q ⊆ Q(ζ4 ) ⊆ Q(ζ8 ). This is because ζ82 is a 4th root of unity. (For example,
( )2
m/m′
1+i
√ = i.) More generally, if m′ | m, then ζm is an m′ th root of unity, so Q(ζm
′
)⊆
2 ∏
Q(ζm ). If m = ℓ mℓ is a prime power factorization, i.e., the mℓ are powers of distinct
primes, then Q(ζmℓ ) ⊆ Q(ζm ), and there is an isomorphism
⊗ ∼
Q(ζmℓ ) −
→ Q(ζm )
ℓ

∏ ∼
such that ⊗ℓ aℓ 7→ ℓ aℓ . For example, 72 = 23 32 , so Q(ζ23 ) ⊗ Q(ζ32 ) − → Q(ζ72 ) via
a ⊗ b 7→ ab. The inclusion Z[ζmℓ ] ,→ Q(ζmℓ ) induces an injective1 ring homomorphism
⊗ ⊗
Z[ζmℓ ] ,→ Q(ζmℓ ),
ℓ ℓ

⊗ ∼
which can be shown to be integral. Hence, ℓ Z[ζmℓ ] −
→ Z[ζm ].

B.2 The Space H and the Canonical Embedding

B.2.1 The Space H
∗
CZm denotes the set of all functions Z∗m → C. It has an obvious ring structure. It can be
viewed as the Cartesian product Cn whose elements are indexed by the elements of Z∗m ,
and both addition and multiplication are component-wise.
The space H is defined by
{ ∗ }
H := x ∈ CZm | xi = x̄m−i ∀i ∈ Z∗m .
1
Recall that an abelian group is flat if and only if torsion-free.

98
 ∗ ∗
H ⊆ CZm is a real subspace of dimension n. In fact, the C-inner product on CZm induces
∼
an R-inner product on H, and there is an R-inner product space isomorphism Rn − →H
via the n-by-n unitary matrix
 
1 i
 ..
..
. 
 . 
1  1 i


U=√  .
2 1 −i 
 
 ..
. . .. 
1 −i

Examples
{ }
• m = 4 : H = (a + ib, a − ib) ∈ C{1,3} | a, b ∈ R ≃ R2 via
( )
1 1 i
U=√ .
2 1 −i

Note that det U = −i.

{ }
• m = 8 : H = (a + id, b + ic, b − ic, a − id) ∈ C{1,3,5,7} | a, b, c, d ∈ R ≃ R4 via
 
1 0 0 i
1  0 1 i 0 
U=√ 
.
2 0 1 −i 0 
1 0 0 −i

Note that det U = −1.

Lattice in H By a lattice in H, we will mean the image of a full-rank lattice in Rn

∼
under the isomorphism Rn −→ H. Equivalently, it is the free abelian group generated
by an R-basis of H. If L ⊆ H is a lattice generated by an R-basis b0 , . . . , bn−1 , its
determinant is defined by
det L := | det B|,
where B = [b0 · · · bn−1 ]. If c0 , . . . , cn−1 ∈ H is another R-basis generating L, then B = CV
for some unimodular matrix V , so det L is independent of the choice of an R-basis of H
generating L.

B.2.2 The Canonical Embedding

Let i ∈ Z∗m , and ωm some fixed mth root of unity. By the universal property of polynomial
rings, the inclusion Q ,→ C extends uniquely to a ring homomorphism

Q[x] → C

such that x 7→ ωm i i
. The kernel is generated by the minimal polynomial of ωm , which
is Φm (x). Hence, there is an injective Q-algebra homomorphism Q[x]/Φm (x) ,→ C such
that x̄ 7→ ωm
i
, i.e.,
σi : Q(ζm ) ,→ C

99
such that ζm 7→ ωm
i
. σi are none other that the n Galois automorphisms on Q(ζm ) fixing
Q. In particular, they are independent, up to a permutation, of the choices of ζm and
ωm . Since R = Z[ζm ], σi is also an automorphism on R fixing Z.
The canonical embedding is the function
∗
σ : Q(ζm ) → CZm , a 7→ (σi (a))i∈Z∗m .
It is an injective Q-algebra homomorphism. Since ωm
m−i i
= ω̄m , the image of σ lies in H.

Lattice in H induced by a Q-basis of Q(ζm ) If x0 , . . . , xn−1 ∈ Q(ζm ) is a Q-basis,

σ may be represented by the n-by-n matrix (σi (xj )). Note that the jth column of this
matrix is σ(xj ) ∈ H. For the power basis 1, ζm , . . . , ζm
n−1
, the matrix becomes
j
S := (σi (ζm )).
Note that a different choice of ζm or ωm results in a permutation of rows or columns of
S. Each σi induces a character Q(ζm )∗ → C∗ , so by the independence of characters, the
rows of (σi (xj )) are linearly independent over C. Hence, σ(x0 ), . . . , σ(xn−1 ) generate a
lattice in H with determinant | det σi (xj )| ̸= 0.
It follows that if G ⊆ Q(ζm ) is a free abelian subgroup with a basis g0 , . . . , gn−1 , then
σ(G) is a lattice in H generated by σ(g0 ), . . . , σ(gn−1 ), and det σ(G) = | det σi (gj )| ̸= 0.
In particular, σ(R) is a lattice in H, and det σ(R) = | det S| ̸= 0. ∑
If x0 , . . . , xn−1 ∈ Q(ζm ) are linearly dependent over Q, then j qj xj = 0 for some
qj ∈ Q, not all zero, and since σi is Q-linear,
( )
∑ ∑
qj σi (xj ) = σi qj xj = 0.
j j

Hence, the columns of the matrix (σi (xj )) are linearly dependent over Q, so det σi (xj ) = 0.
Note the following:
∑
• If x0 , . . . , xn−1 , y0 , . . . , yn−1 ∈ Q(ζm ) and yj = k Mjk xk , where Mjk ∈ Q, then
det σi (yj ) = (det M )(det σi (xj )).

• If x0 , . . . , xn−1 , y ∈ Q(ζm ), then

( )
∏
det σi (yxj ) = det σi (y)σi (xj ) = σi (y) det σi (xj ) = N(y) det σi (xj ),
i

Examples
• m=4: ( ) ( )
σ1 (1) σ1 (ζ4 ) 1 i
S= =
σ3 (1) σ3 (ζ4 ) 1 −i
det S = −2i, so det σ(R) = | − 2i| = 2.
• m=8:
   
σ1 (1) σ1 (ζ8 ) σ1 (ζ82 ) σ1 (ζ83 ) 1 ζ8 ζ82 ζ83
 σ3 (1) σ3 (ζ8 ) σ3 (ζ82 ) σ3 (ζ83 )   ζ83 −ζ82 
S=   1
3  = 
ζ8 
 σ5 (1) σ5 (ζ8 ) σ5 (ζ82 ) σ5 (ζ8 ) 1 −ζ8 ζ82 −ζ83 
σ7 (1) σ7 (ζ8 ) σ7 (ζ82 ) σ7 (ζ83 ) 1 −ζ83 −ζ82 −ζ8
det S = −16, so det σ(R) = | − 16| = 16.

100
Trace and norm For a ∈ Q(ζm ), define
∑ ∏
Tr(a) := σi (a), N(a) := σi (a).
i∈Z∗m i∈Z∗m

Note that if a ∈ R, then Tr(a) ∈ R and N(a) ∈ R.

Note the following:
• Tr(a) and N(a) are independent of the choices of ζm and ωm .
• Tr(1) = n and N(1) = 1.
• Tr is Q-linear.
• N(ab) = N(a) N(b) for all a, b ∈ Q(ζm ). Hence, if u ∈ R is a unit, then N(u) = ±1.
(The converse is also true; see Corollary B.2.3.)
• For all a ∈ Q(ζm ), N(a) = 0 if and only if a = 0.
Proposition B.2.1. Let a ∈ Q(ζm ), and A an n-by-n matrix with entries in Q rep-
a
resenting the multiplication map Q(ζm ) −
→ Q(ζm ) with respect to some Q-basis. Then
Tr A = Tr(a) and det A = N(a).
Proof. Let f = det(xI − A) be the characteristic polynomial. Then clearly

f = xn − (Tr A)xn−1 + · · · + (−1)n det A. (B.1)

By the Cayley-Hamilton theorem, a is a root of f , so f is divisible by the minimal

polynomial of a. In fact, it is easy to show that if ma ∈ Q[x] is the minimal polynomial
n/d
of a with degree d, then f = ma .
Let a0 , . . . , ad−1 ∈ Q(ζm ) be the roots of ma (they all lie in Q(ζm ) since Q ⊆ Q(ζm )
is a splitting field extension of Φm , hence normal), so that ma = (x − a0 ) · · · (x − ad−1 ).
Then ( )n/d
n∑ ∏
ma = x −
n/d n
ai xn−1
+ · · · + (−1) n
ai ,
d i i

so comparing with (B.1), we see that

( )n/d
n∑ ∏
Tr A = ai , det A = ai . (B.2)
d i i

Now consider the extensions

d n/d
Q ⊆ Q(a) ⊆ Q(ζm ).

Being separable, a0 , . . . , ad−1 ∈ Q(ζm ) are distinct, and there are exactly d embeddings
of Q(a) into Q(ζm ) fixing Q (corresponding to a 7→ ai ), each of which extends to exactly
n/d automorphisms (Q ⊆ Q(ζm ) being normal) of Q(ζm ) fixing Q. It follows that
( )n/d
∑ n∑ ∏ ∏
σi (a) = ai , σi (a) = ai .
i∈Z∗
d i i∈Z∗ i
m m

By (B.2), it follows that Tr(a) = Tr A and N(a) = det A.

101
Corollary B.2.2. i. If a ∈ Q(ζm ), then Tr(a) ∈ Q and N(a) ∈ Q.

ii. If a ∈ R, then Tr(a) ∈ Z and N(a) ∈ Z.

Proof. i is immediate from Proposition B.2.1. If a ∈ R, then

Tr(a), N(a) ∈ R ∩ Q = Z,

since Z is integrally closed.

Alternatively, any Z-basis of R is a Q-basis of Q(ζm ), and with respect to this basis,
r r
the multiplication maps Q(ζm ) − → Q(ζm ) and R − → R by r ∈ R are represented by the
same matrix A with integer entries, whoses trace and determinant are integers.

Corollary B.2.3. For all r ∈ R, N(r) = ±1 if and only if r is a unit in R.

Proof. Any Z-basis of R is a Q-basis of Q(ζm ), and with respect to this basis, the mul-
r r
tiplication maps Q(ζm ) − → Q(ζm ) and φ : R − → R are represented by the same matrix A
with integer entries. Now r is a unit in R if and only if φ is an isomorphism if and only
if det A ∈ Z is a unit, i.e., det A = ±1. Since det A = N(r), the result follows.

Proposition B.2.4. If x0 , . . . , xn−1 ∈ Q(ζm ), then det Tr(xi xj ) = (det σi (xj ))2 .

Proof. ∑ ∑
Tr(xi xj ) = σk (xi xj ) = σk (xi )σk (xj ) = (AT A)ij ,
k∈Z∗m k∈Z∗m

where A is the matrix (σi (xj )). Hence, det Tr(xi xj ) = (det A)2 .
∑
Remark B.2.5. Note that If x0 , . . . , xn−1 , y0 , . . . , yn−1 ∈ Q(ζm ) and yj = k Mjk xk ,
where Mjk ∈ Q, then
det Tr(yi yj ) = (det M )2 det Tr(xi xj ).

Corollary B.2.6. If b0 , . . . , bn−1 is a basis of a free abelian subgroup G ⊆ Q(ζm ), then

det σ(G) = | det Tr(bi bj )|.

Corollary B.2.7. i. If x0 , . . . , xn−1 ∈ Q(ζm ), then (det σi (xj ))2 ∈ Q.

ii. If r0 , . . . , rn−1 ∈ R, then (det σi (rj ))2 ∈ Z. In particular, (det S)2 ∈ Z.

Corollary B.2.8. x0 , . . . , xn−1 ∈ Q(ζm ) is a Q-basis if and only if det Tr(xi xj ) ̸= 0.

Examples

• m = 4 : If ω4 = i, then

σ : Q(ζ4 ) → C{1,3} ,
q0 + q1 ζ4 7→ (q0 + iq, q0 − iq1 )

where q0 , q1 ∈ Q. Hence,
Tr(q0 + q1 ζ4 ) = 2q0 ∈ Q,
N(q0 + q1 ζ4 ) = q02 + q12 ∈ Q.

102
 Since ζ42 = −1,

(p0 + p1 ζ4 )(q0 + q1 ζ4 ) = p0 q0 − p1 q1 + (p0 q1 + p1 q0 )ζ4 ,

where p0 , p1 , q0 , q1 ∈ Q. Hence, in terms of the basis 1, ζ4 , the multiplication map

q0 +q1 ζ4
Q(ζ4 ) −−−−→ Q(ζ4 )

is represented by the matrix

( )
q0 −q1
A= .
q1 q0

Note that
Tr A = 2q0 = Tr(q0 + q1 ζ4 ),
det A = q02 + q12 = N(q0 + q1 ζ4 ).

• m = 8 : If ω8 = 1+i
√ ,
2
then

σ : Q(ζ8 ) → C{1,3,5,7} ,
q0 + q1 ζ8 + q2 ζ82 + q3 ζ83 7→ (q0 + q1 ζ8 + q2 ζ82 + q3 ζ83 ,
q0 + q3 ζ8 − q2 ζ82 + q1 ζ83 ,
q0 − q1 ζ8 + q2 ζ82 − q3 ζ83 ,
q0 − q3 ζ8 − q2 ζ82 − q1 ζ83 )

where q0 , q1 , q2 , q3 ∈ Q. Hence,

Tr(q0 + q1 ζ8 + q2 ζ82 + q3 ζ83 ) = 4q0 ∈ Q,

and after some calculation, one can verify that

N(q0 + q1 ζ8 + q2 ζ82 + ζ83 ) = (q02 − q22 + 2q1 q3 )2 + (q12 − q32 − 2q0 q2 )2 ∈ Q.

It is easy to see that in terms of the basis 1, ζ8 , ζ82 , ζ83 , the multiplication map
q0 +q1 ζ8 +q2 ζ 2 +q3 ζ 3
Q(ζ8 ) −−−−−−−−8−−−→
8
Q(ζ8 )

is represented by the anti-circulant matrix

 
q0 −q3 −q2 −q1
 q1 q0 −q3 −q2 
A=  q2
.
q1 q0 −q3 
q3 q2 q1 q0

Note that
Tr A = 4q0 = Tr(q0 + q1 ζ8 + q2 ζ82 + q3 ζ83 ).
One can also verify that

det A = N(q0 + q1 ζ8 + q2 ζ82 + q3 ζ83 ).

103
B.3 Discriminant
The discriminant of Q(ζm ) is defined by
∆Q(ζm ) := (det σ(R))2 .
Hence, if b0 , . . . , bn−1 ∈ R is any integral basis, then
∆Q(ζm ) = | det σi (bj )|2 = | det Tr(bi bj )|.
In particular,
∆Q(ζm ) = | det S|2 .
Since (det S)2 ∈ Z (Corollary B.2.7), ∆Q(ζm ) is a positive integer.2
∑
If x0 , . . . , xn−1 ∈ Q(ζm ), then xj = i Mji bi for some Mji ∈ Q, so (see Remark B.2.5)
| det Tr(xi xj )| = (det M )2 ∆Q(ζm ) . (B.3)

j
Relationship with polynomial discriminant Since S = (σi (ζm )) is a Vandermonde
matrix, ∏ ∏
det S = (σi (ζm ) − σj (ζm )) = i
(ωm − ωm
j
).
i<j i<j
∏
Hence, in terms of ∆Φm := i
i<j (ωm − j 2
ωm ), we have
∆Q(ζm ) = |∆Φm |.

Examples
• m = 4 : We know that det S = −2i, so ∆Q(ζm ) = | det S|2 = 4. On the other hand,
( )
i j Tr(ζ40 ζ40 ) Tr(ζ40 ζ41 )
Tr(ζ4 ζ4 ) =
Tr(ζ41 ζ40 ) Tr(ζ41 ζ41 )
( )
Tr(1) Tr(i)
=
Tr(i) Tr(−1)
( )
1+1 i−i
=
i − i −1 − 1
( )
2 0
= ,
0 −2
so
∆Q(ζm ) = | det Tr(ζ4i ζ4j )| = | − 4| = 4.

• m = 8 : We know that det S = −16, so ∆Q(ζm ) = | det S|2 = 28 . On the other

hand,
   
Tr(ζ80 ζ80 ) Tr(ζ80 ζ81 ) Tr(ζ80 ζ82 ) Tr(ζ80 ζ83 ) 4 0 0 0
 Tr(ζ81 ζ80 ) Tr(ζ81 ζ81 ) Tr(ζ81 ζ82 ) Tr(ζ81 ζ83 )   0 −4 
Tr(ζ8i ζ8j ) =    0
2 3  = 
0 ,
 Tr(ζ82 ζ80 ) Tr(ζ82 ζ81 ) Tr(ζ82 ζ82 ) Tr(ζ8 ζ8 ) 0 0 −4 0 
Tr(ζ83 ζ80 ) Tr(ζ83 ζ81 ) Tr(ζ83 ζ82 ) Tr(ζ83 ζ83 ) 0 −4 0 0
so
∆Q(ζm ) = | det Tr(ζ8i ζ8j )| = |44 | = 28 .
2
A more standard definition of ∆Q(ζm ) is det S, which can be negative.

104
B.4 Ideals
If I ⊆ R is an ideal, define N(I) := |R/I|. Note that N(I) ≥ 1, where equality holds if
and only if I = R.

Proposition B.4.1. If I ⊆ R is a nonzero ideal, then N(I) is finite.

Proof. Let I ⊆ R be a nonzero ideal. Then there exists a nonzero element a ∈ I, and
∏
N(a) = a σi (a) ̸= 0.
i∈Z∗m ,
ωmi ̸=ζ
m

∏
Since a ∈ R, σi (a) ∈ R for all i ∈ Z∗m , so ζm ̸=ωm
i σi (a) ∈ R. Since a ∈ I, it follows that

N(a) ∈ I. Since a ∈ R, N(a) ∈ Z. Now R ≃ Zn as an abelian group, so R/ N(a)R ≃

(Z/ N(a)Z)n as an abelian group. Since N(a) ̸= 0, |(Z/ N(a)Z)n | = n N(a) < ∞, i.e.,
|R/ N(a)R| is finite. Since N(a)R ⊆ I ⊆ R and R/I ≃ (R/ N(a)R)/(I/ N(a)R), |R/I|
must be finite, too.

Remark B.4.2. It follows that N(I) is a positive integer for every nonzero ideal I ⊆ R.

Corollary B.4.3. If I ⊆ R is a nonzero ideal, then N(I) ∈ I.

Proof. Let N(I) = k ∈ Z. Since |R/I| = k, for 1̄ ∈ R/I, we must have k · 1̄ = 0 ∈ R/I,
i.e., k ∈ I.

Corollary B.4.4. Every nonzero ideal of R is a free abelian group of rank n.

Remark B.4.5. It follows that if I ⊆ R is a nonzero ideal, then σ(I) is a lattice in H.

Proposition B.4.6. If b1 , . . . , bk ∈ R are linearly independent over Z and r ∈ R is a

nonzero element, then rb1 , . . . , rbk are linearly independent over Z.

Proof. Consider the equation

0 = c1 rb1 + · · · + ck rbk = r(c1 b1 + · · · + ck bk ),

where c1 , . . . , ck ∈ Z. In view of R = Z[x]/Φm (x), we may represent r and bi as poly-

nomials in Z[x] of degree less than n, say r = f¯ and bi = ḡi . Then the equation above
implies that Φm (x) divides f (c1 g1 + · · · + ck gk ) in Z[x]. Since Φm (x) is irreducible, it is
a prime (Z[x] being a UFD), so it divides f or c1 g1 + · · · + ck gk . Since the degrees of f
and c1 g1 + · · · + ck gk are both less than n, we must have c1 g1 + · · · + ck gk = 0. Then
c1 b1 + · · · + ck bk = 0, so c1 = · · · = ck = 0 by the linear independence of b1 , . . . , bk .

Corollary B.4.7. If b0 , . . . , bn−1 ∈ R is an integral basis and r ∈ R is a nonzero element,

then the ideal ⟨r⟩ ⊆ R is a free abelian group with a basis rb0 , . . . , rbn−1 .

Lemma B.4.8. Let b0 , . . . , bn−1 ∑ ∈ R be an integral basis, I ⊆ R a nonzero ideal with a

Z-basis c0 , . . . , cn−1 , and cj = i Mji bi , where Mji ∈ Z. Then

N(I) = | det M |.

105
Proof. There exists an integral basis b′0 , . . . , b′n−1 ∈ R such that k0 b′0 , . . . , kn−1 b′n−1 is a
Z-basis of I for some k0 , . . . , kn−1 ∈ Z. Then clearly N(I) = |k0 · · · kn−1 |, so N(I) =
| det M ′ |, where ′
∑ M′ is the n-by-n diagonal matrix with diagonal entris k′ 0 , . . . , kn−1 , so
that kj bj = i Mji bi . More generally, change of bases corresponds to M 7→ U M ′ V for
some unimodular matrices U and V , so | det M ′ | remains unchanged.

Proposition B.4.9. If I ⊆ R is a nonzero ideal, then

(det σ(I))2 = N(I)2 ∆Q(ζm ) .

Proof.∑Let b0 , . . . , bn−1 ∈ R be an integral basis, and c0 , . . . , cn−1 a Z-basis of I. Then

cj = i Mji bi for some Mji ∈ Z, so by (B.3) and Lemma B.4.8,

| det Tr(ci cj )| = N(I)2 ∆Q(ζm ) .

Hence,
(det σ(I))2 = | det σi (cj )|2 = | det Tr(ci cj )| = N(I)2 ∆Q(ζm ) .

Proposition B.4.10. If r ∈ R is a nonzero element, then N(⟨r⟩) = | N(r)|.

Proof. Let b0 , . . . , bn−1 ∈ R be an integral basis. By Corollary B.4.7, rb0 , . . . , rbn−1 is a

Z-basis of ⟨r⟩, so by Proposition B.4.9,

| det σi (rbj )|2 = N(⟨r⟩)2 ∆Q(ζm ) . (B.4)

On the other hand, det σi (rbj ) = N(r) det σi (bj ), so

| det σi (rbj )|2 = N(r)2 | det σi (bj )|2 = N(r)2 ∆Q(ζm ) . (B.5)

Since ∆Q(ζm ) ̸= 0, (B.4) and (B.5) gives N(⟨r⟩) = | N(r)|.

Remark B.4.11. Note that the equality does not hold if r = 0.

Lemma B.4.12. If I ⊆ R is a nonzero ideal and P ⊆ R is a nonzero prime ideal, then

there is a ring isomorphism I/P I ≃ R/P .

Proof. Since R is a Dedekind domain, P I and I are distinct ideals, and there is no ideal
between P I and I. Hence, I/P I is an R/P -module with no intermediate submodule, so
it is generated by a single nonzero element. Since P ⊆ R is a maximal ideal, this means
that I/P I ≃ R/P .

Proposition B.4.13. N(IJ) = N(I) N(J) for all ideals I, J ⊆ R.

Proof. If I = 0, then IJ = 0, so N(IJ) = N(I) = |R|, so the equality becomes

|R| = |R| · N(J).

Since R is infinite and N(J) ≤ |R|, the equality does hold.

Now assume that I ̸= 0 and J ̸= 0. Since R is a Dedekind domain, I is a product of
nonzero prime ideals. If it is an empty product, i.e., I = R, then the equality becomes

106
N(J) = 1 · N(J), which is obviously true. So we may assume that I ̸= R, and it suffices
to show that N(P J) = N(P ) N(J) for every nonzero prime ideal P ⊆ R.
From the ring isomorphism
R/P J
≃ R/J,
J/P J
we have |R/P J| = |J/P J| · |R/J|, i.e., N(P J) = |J/P J| · N(J). (Note that all three
quantities are finite by Proposition B.4.1.) By Lemma B.4.12, |J/P J| = |R/P | = N(P ),
so N(P J) = N(P ) N(J), as desired.

Corollary B.4.14. If I, J ⊆ R are nonzero ideals, then |I/IJ| = N(J).

Proof. From the ring isomorphism

R/IJ
≃ R/I,
I/IJ

we have |R/IJ| = |R/I| · |I/IJ|, i.e., N(IJ) = N(I) · |I/IJ|. On the other hand, N(IJ) =
N(I) N(J) by Proposition B.4.13. Since all quantities here are finite and N(I) ̸= 0, we
have |I/IJ| = N(J).

Corollary B.4.15. Let I ⊆ R be a nonzero ideal. If N(I) is prime, then I is a prime

ideal.

Proof. If I = R, then N(I) = 0 is not prime, so I ̸= R. Hence, I is a product of

at least one prime ideal. Suppose that I = J1 J2 , where J1 , J2 ⊆ R are ideals. Then
N(I) = N(J1 ) N(J2 ), and since N(I) is prime, N(J1 ) = 1 or N(J2 ) = 1, i.e., J1 = R or
J2 = R. This shows that I is a product of at most one prime ideal.

B.4.1 Fractional ideals

An R-submodule I ⊆ Q(ζm ) is called a fractional ideal of R if there exists a nonzero
d ∈ R such that dI ⊆ R. Note that every ideal of R is a fractional ideal. An ideal of R
is sometimes called an integral ideal. Since every nonzero integral ideal is a free abelian
group of rank n (Corollary B.4.4), so is every nonzero fractional ideal. It follows that if
I ⊆ Q(ζm ) is a nonzero fractional ideal, then σ(I) is a lattice in H.
Note the following:

• If I ⊆ Q(ζm ) is a fractional ideal such that dI ⊆ R for some nonzero d ∈ R, then

b0 , . . . , bn−1 is a Z-basis of I if and only if db0 , . . . , dbn−1 is a Z-basis of the integral
ideal dI.

• Since Q(ζm ) is a field of fractions for R, every finitely generated R-submodule of

Q(ζm ) is a fractional ideal of R. In particular, every principal R-submodule of
Q(ζm ) is a fractional ideal.

• If I, J ⊆ Q(ζm ) are fractional ideals, then so are I + J and IJ.

The norm of a nonzero fractional ideal I ⊆ Q(ζm ) is defined by

N(I) := N(dI)/| N(d)| ∈ Q,

107
where d ∈ R is a nonzero element such that dI ⊆ R. This is well-defined: if e ∈ R
is another nonzero element such that eI ⊆ R, then by Proposition B.4.10 and Proposi-
tion B.4.13,

| N(e)| N(dI) = N(⟨e⟩) N(dI) = N(edI) = N(deI) = N(⟨d⟩) N(eI) = | N(d)| N(eI).

For nonzero integral ideals, this definition of norm agrees with the earlier definition
of norm. Note that the norm of a nonzero fractional ideal is a positive rational number.

Proposition B.4.16. Let I ⊆ Q(ζm ) be any subset. TFAE:

i. I is a nonzero fractional ideal of R.

ii. There exists d ∈ R such that dI is a nonzero ideal of R.

Proof. i⇒ii: By definition, there exists a nonzero d ∈ R such that dI ⊆ R. Since

I ⊆ Q(ζm ) is an R-submodule, so is dI, i.e., dI ⊆ R is an ideal. Since d ̸= 0 and I ̸= 0,
dI ̸= 0
i⇐ii: Since dI ⊆ R is a nonzero ideal by assumption, d ̸= 0 and I ̸= 0. Hence, all we
have to show is that I ⊆ Q(ζm ) is an R-submodule. First note that since dI ⊆ R is an
ideal, I ̸= ∅.

• Let i1 , i2 ∈ I. Since dI is an abelian group, di1 + di2 ∈ dI, i.e., d(i1 + i2 ) ∈ dI.
Since d ̸= 0, this implies that i1 + i2 ∈ I.

• Let i ∈ I and r ∈ R. Since dI ⊆ R is an ideal, rdi ∈ dI, i.e., d(ri) ∈ dI. Since
d ̸= 0, this implies that ri ∈ I.

Hence, I ⊆ Q(ζm ) is an R-submodule.

Proposition B.4.17. If x ∈ Q(ζm ) is a nonzero element, then N(⟨x⟩) = | N(x)|.

Proof. Since Q(ζm ) is a field of fractions for R, dx ∈ R for some nonzero d ∈ R. Then
d⟨x⟩ ⊆ R, so

N(d⟨x⟩) N(⟨dx⟩) | N(dx)| | N(d) N(x)|

N(⟨x⟩) = = = = = |N (x)|.
| N(d)| | N(d)| | N(d)| | N(d)|

Proposition B.4.18. If I, J ⊆ Q(ζm ) are fractional ideals, then N(IJ) = N(I) N(J).

Proof. If I = 0, then IJ = 0 and N(IJ) = N(I) = |R|, so the equality becomes

|R| = |R| · N(J).

Since R is infinite and N(J) ≤ |R|, the equality does hold.

Now assume that I ̸= 0 and J ̸= 0. If d, e ∈ R are nonzero elements such that
dI, eJ ⊆ R, then deIJ ⊆ R, so
N(deIJ) N(dI) N(eJ) N(dI) N(eJ)
N(IJ) = = = · = N(I) N(J).
| N(de)| | N(d) N(e)| | N(d)| | N(e)|

108
Proposition B.4.19. If I ⊆ Q(ζm ) is a nonzero fractional ideal, then

(det σ(I))2 = N(I)2 ∆Q(ζm ) .

Proof. Let dI ⊆ R, where d ∈ R. By Proposition B.4.9,

N(dI)2 (det σ(dI))2

N(I)2 ∆Q(ζm ) = ∆ = .
| N(d)|2 Q(ζm ) | N(d)|2

Let b0 , . . . , bn−1 be a Z-basis of I. Then db0 , . . . , dbn−1 is a Z-basis of dI, so

det σ(dI) = | det σi (dbj )| = | N(d)| · | det σi (bj )| = | N(d)| det σ(I).

Hence, N(I)2 ∆Q(ζm ) = (det σ(I))2 .

Acknowledgments
This research was supported by Fusion Technology R&D Center of SK Telecom and
UNIST.

109
110
Bibliography

[ACPS09] B. Applebaum, D. Cash, C. Peikert, and A. Sahai. Fast cryptographic primi-

tives and circularsecure encryption based on hard learning problems. In CRYPTO,
595–618. 2009.

[Bab85] L. Babai. On Lovasz lattice reduction and the nearest lattice point problem.
Combinatorica, 6(1):1–13, 1986. Preliminary version in STACS 1985.

[Ban93] Z. Brakerski, C. Gentry, and V. Vaikuntanathan. (Leveled) fully homomorphic

encryption without bootstrapping. In ICTS, 309–325. 2012.

[Bar12] Bar-Ilan Univ. Winter School on Lattice-Based Cryptography and Applications.

2012.

[Cam14] P. Campbell, M. Groves, and D. Shepherd. Soliloquy: A Cautionary Tale, 2014.

http://docbox.etsi.org/Workshop/2014/201410_CRYPTO/S07_Systems_and_
Attacks/S07_Groves_Annex.pdf

[Con09] K. Conrad. The Different Ideal.

http://www.math.uconn.edu/~kconrad/blurbs/gradnumthy/different.pdf

[Cra15] R. Cramer, L. Ducas, C. Peikert, O. Regev. Recovering Short Generators of

Principal Ideals in Cyclotomic Rings, 2015.
https://eprint.iacr.org/2015/313.pdf

[Ding07] J. Ding and Richard Lindner. Identifying Ideal Lattices, 2007.

https://eprint.iacr.org/2007/322

[GPV08] C. Gentry, C. Peikert, and V. Vaikuntanathan. Trapdoors for hard lattices and
new cryptographic constructions. In STOC, pages 197–206. 2008.

[Hu15] Yupu Hu and Huiwen Jia. Cryptanalysis of GGH Map, 2015.

https://eprint.iacr.org/2015/301.pdf

[LPR10] V. Lyubashevsky, C. Peikert, and O. Regev. On ideal lattices and learning with
errors over rings. J. ACM, 2013. To appear. Preliminary version in Eurocrypt 2010.

[LPR13] Vadim Lyubashevsky, Chris Peikert, and Oded Regev. A toolkit for ring-LWE
cryptography. In EUROCRYPT, 35–54. 2013.

[LSS14] Adeline Langlois, Damien Stehle, Ron Steinfeld. GGHLite: More Efficient Mul-
tilinear Maps from Ideal Lattices. In EUROCRYPT, 239–256. 2014.

111
[MP12] D. Micciancio and C. Peikert. Trapdoors for lattices: Simpler, tighter, faster,
smaller. In EUROCRYPT, 700–718. 2012.

[MR04] D. Micciancio and O. Regev. Worst-case to average-case reductions based on

Gaussian measures. SIAM J. Comput., 37(1):267–302, 2007. Preliminary version in
FOCS 2004.

[P13] C. Peikert. Tutorials from crypt@b-it 2013 summer school at Bonn University.
http://www.cc.gatech.edu/~cpeikert, 2013.

[Reg05] O. Regev. On lattices, learning with errors, random linear codes, and cryptog-
raphy. J. ACM, 56(6):1–40, 2009. Preliminary version in STOC 2005.

[Sma09] N.P. Smart and F. Vercauteren. Fully Homomorphic Encryption with Relatively
Small Key and Ciphertext Sizes, 2009.
https://eprint.iacr.org/2009/571

112