See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/268214316

An Implementation of AES Algorithm in FPGA

Conference Paper · September 2014

CITATION READS

1 327

4 authors:

Isaac Nattan da Silva Palmeira Alcir Cledson de S. Góis

Universidade Federal de Sergipe Universidade Federal de Sergipe
3 PUBLICATIONS   5 CITATIONS    1 PUBLICATION   1 CITATION   

SEE PROFILE SEE PROFILE

Wanderson Roger Azevedo Dias Edward David Moreno Ordonez

Instituto Federal Sergipe Universidade Federal de Sergipe
34 PUBLICATIONS   54 CITATIONS    133 PUBLICATIONS   153 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

security embedded View project

Security in Embedded Systems with SIP Protocol View project

All content following this page was uploaded by Wanderson Roger Azevedo Dias on 13 November 2014.

The user has requested enhancement of the downloaded file.

 An Implementation of AES Algorithm in FPGA
Isaac Nattan da Silva Palmeira, Alcir Cledson de S. Góis, Wanderson Roger Azevedo Dias,
Edward David Moreno
Department of Computing - DComp
Federal University of Sergipe - UFS
Aracaju, Sergipe, Brazil
{isaacnattan2, wradias, edwdavid}@gmail.com, [email protected]

ABSTRACT The size of the array will depend on the block size used,
This article aims to present an alternative implementation of the consisting of four rows and columns Nb, where Nb is the
Rijndael algorithm, the AES (Advanced Encription Standart). number of bits divided by the block 32. The proposed algorithm
The algorithm described above is able to encrypt pieces of 16- has the ability to encrypt text of 128 bits (16 bytes) , thus the
byte text using a key of the same size. The basic operations of state will have 4x4 size. The main key is grouped in the same
the AES operation will be described: AddRoundKey, SubBytes, way that the state with Nk columns.
ShiftRows, MixColumns, and their respective inverses still a As the number of rounds AES or number of rounds varies
key generator algorithm (KeyExpansion). depending on the key length, and Nr (number of rounds) of 10,
12 and 14 for Nk (number of columns) equal to 4, 6 and 8
Keywords respectively. In each round of the encryption algorithm, five
cryptography; AES algorithm; encryption; decryption. steps are performed: AddRoundKey, SubBytes, ShiftRows,
MixColumns and KeyExpansion algorithm. In the last round,
however, the MixColumns operation is suppressed.
1. INTRODUCTION
Encryption originates from the greek word which means cryptos
secret or hidden. The objective of cryptography are encryption 2.1 Mathematical Considerations
methods or data messages so that only the legitimate receiver of Various operations are defined in the AES bytes, with one byte
certain information may have access to it. In the treatment of represented in the finite field GF (28), which mathematically
problems related to information security is added to it the defines a Galois field, also known as Galois Field (GF).
existence of cryptanalysis, responsible for studying the means
of deciphering an encrypted message without all the 1.1.1. GF(28)
information needed to decode the message correctly. The GF (28) symbolizes a field, algebraic structure which has
two operations {+, ∙} addition and multiplication respectively,
Encryption electronics began to be used after the Second World
closed on elements of the field.
War. But not until 1974 that the first cryptographic algorithm
has been used in a commercial manner. Lucifer was developed Definition 1.1: a field is a set F with two laws of composition C
by IBM and, after several changes made by the NSA (National and such that.
Security Agency), came to be called DES (Data Encryption
Standart) then being used as a U.S. cryptographic standard.  (F +) / C is a commutative group;
 (F ∙) /, where F \ {0}, is a commutative group;
For 20 years, DES was the default algorithm used by the  the distributive law holds.
Yankee government to protect confidential information. The
emergence of the AES (Advanced Encryption Standard) was A field contains at least two distinct elements '0 'and '1'.
due to the great need to replace DES, which has become
outdated because of the small key size (56 bits) used. For this, 2.1.2. Addition in GF(28)
the NIST (National Institute of Standards and Technology) has The addition of polynomials in GF (28), corresponds to the
launched a competition in 1997 to adopt the new symmetric XOR (exclusive OR) bitwise.
algorithm, which would be called AES.
The algorithm should meet some requirements such as: 2.1.3. Multiplication in GF(28)
copyright free, public disclosure, faster compared to DES, block In the polynomial representation, this procedure corresponds to
cipher with 128-bit keys of 128, 192 and 256 bits, possibility of polynomial multiplication already known, ie, applying the
implementation in software and hardware. distributive property, whereas the degree of the resulting
polynomial can never be greater than or equal to eight, hence
The algorithm was created by Belgian Vincent Rijmen and Joan the name GF (28). For the degree of the polynomial higher than
Daemen. As both the first in the second stage of the competition or equal to eight (8 ≥ g) is added to the irreducible polynomial
all the algorithms described meeting requirements of the tender, m(x)= xg + xg-4 + xg-5 + xg-7 + xg-8.
the decision was made based on other qualities such as security,
flexibility, good performance in software and hardware etc.
2.2 AES Transformations
Possession of some mathematical considerations and a brief
2. ALGORITHMIC STRUCTURE OF AES introduction of algorithmic structure of the AES, the basic
Some concepts are of vital importance for the understanding of functions that operate behind the encryption/decryption of a
the proposed algorithm, one is understanding what comes to the block process will be presented.
state. The state is an array of bytes that will be handled during
the various rounds or rounds, and therefore will change every
step.
2.2.1. AddRoundKey
The AddRoundKey operation is nothing more than a bitwise
xor operation between the state and the key round. This is the
one that has no reverse, in the deciphering process is used with
the following keys generated backwards. The Figure 1 shows
schematically how this operation works.
Figure 3. MixColumns constant Matrix.

In its reverse process happens analogously differing only in the

constant matrix used (see Figure 4).

Figure 4. InvMixColumns constant Matrix.

3. KEYEXPANSION ALGORITHM
For each round in the implementation of the algorithm is
applied to a key, specifically the implementation of the
Figure 1. Operation schematic AddRoundKey. transformation AddRoundKey. These keys are generated by
KeyExpansion algorithm.
The key generation operation is to be applied to each word as
2.2.2. SubBytes and its Inverse they are known the columns of the state, one of the approaches
SubBytes transformation modifies the values of the state based to two distinct situations:
on a substitution box (S-Box), using the bytes from the current
state as indices to the values contained in the S-Box used. The If the first byte of the word is not operated multiple of Nk, the
S-Box is an array of size 16x16 with different hex values. The new word will receive the result of the xor operation between
replacement is performed as follows: the first and second the word and the word immediately preceding the previous key
current number hexadecimal value respectively represent the corresponding position.
row and column of the value contained in the S-Box. In this
If the identifier word op is a multiple of Nk, one xor should be
case the inverse transform process applies, however considering
applied between the replacement using the S-Box, the word
the inverse S-box.
immediately preceding rotated to the left, and the constant
round of RC [j]. Then a new xor between the result of the
2.2.3. ShiftRows and its Inverse previous operation and the word of the previous key
This transformation consists in a rotation left the state lines so: corresponding position will be applied. The following table
(i) the first line is not amended; (ii) the second one suffers a shows all the constants used in the rounds of the algorithm (see
rotation line; (iii) the third line suffers two rotations; (iv) the Table 1).
fourth line suffers three rotations. The Figure 2 shows how this
process works.
Table 1. Constants of Rounds
2.2.4. MixColumns and its Inverse RC[1] 01 ( 0000 0001 )
MixColumns transformation is performed in a matrix RC[2] 02 ( 0000 0010 )
RC[3] 04 ( 0000 0100 )
multiplication in GF (28). The state is being modified
RC[4] 08 ( 0000 1000 )
multiplied by a constant matrix formed by variations of the
RC[5] 10 ( 0001 0000 )
polynomial a (x) = {02}+{03}+{01}+{01}, the matrix formed RC[6] 20 ( 0010 0000 )
by the variations of the polynomial is shown the Figure 3. RC[7] 40 ( 0100 0000 )
RC[8] 80 ( 1000 0000 )
RC[9] 1b ( 0001 1011 )
RC[10] 36 ( 0011 0110 )

4. COMPLETE PROCEDURE OF THE

AES ALGORITHM
Transformations in the encryption process can be followed in
Figure 5. Deciphers In the sequence is exactly the reverse
process to that shown.

Figure 2. Operation schematic ShiftRows.

 Having the data, the algorithm KeyExpansion was the most
costly, but the complexity of the algorithm analyzing your time
will be considered constant compared with the transformation
of the state.
Table 2. Time and Space Statistics
Time Space the FPGA
(ns) (logic elements)
128
AddRoundKey 12.124
<1% of the capacity maximum
6720
SubBytes 19.278
20% of the capacity maximum
64
ShiftRows 12.822
<1% of the capacity maximum
472
MixColumns 17.137
1% of the capacity aximum
8904
KeyExpansion 90.828
27% of the capacity maximum
21421
COMPLETE 771.12
64% of the capacity maximum

Thus to evaluate the method of encryption/decryption simply

collect the transformation of worse weather (SubBytes), this
time is the minimum time clock, and multiply it by the number
Figure 5. Complete procedure of cryptography. of times that each transformation is applied. Thus we have:
Minimum Time = 19.278 ns ∙ 40 = 771.12 ns
The AES algorithm has been implemented with a view to
saving space. Since the goal was to develop a method encrypts With this result it is possible to estimate the throughput as:
and decrypts, we studied the possibility of merging the
Transfer Fee = 128 bits/s 771.12 ≈ 166 Mbits/s
transformations with their inverse and found, from this, the
union between these factors. SubBytes operation for this factor In the spatial aspect, the complete algorithm using totaled
was not found, but this was united with its inverse in order to 21.421 logic elements, about 64% of the maximum capacity of
reduce complexity. the FPGA, among them we can highlight:
ShiftRows transformation was realized that the same rotations  Use of 1.540 bit registers (11 keys, current state and
that its inverse could be applied in reverse order. Thus the round in question);
rotation applied to the third row would be the same for both
cases.  4.186 bits of ROM (S-Box and Inverse, constant
round and constant matrices).
The uniting factor found between the MixColumns operation
and its inverse is the generalization of the constant matrix (see The Figure 7 shows the state machine of the AES.
Figure 6).
Where the variable d assume '0 'and '1 in encryption process' in
the decrypts. Once the changes merged, the complete algorithm
should meet two methods of transformation of the state.
Viewing this way the key generator would have to be
completely executed before this, since in the latter decrypts key
generated would be first processed.

Figure 6. Generalization of constant matrixes. Figure 7. Complete procedure of cryptography.

The transformation of state always start with the AddRoundKey 6. CONCLUSIONS

operation. This would follow in the next transformations, based Compared with other implementations, which separated their
on the method and the round in question. algorithms in cipher and decipher, one can notice the
duplication of all transformations, mainly MixColumns which
5. SOME RESULTS is of considerable size, so there's no doubt about space
All functions were simulated and analyzed in isolation and reduction of the proposed algorithm. With the separation of the
together, with the aid of the tool Amendment Quartus II version methods, times cipher were evaluated separately from the times
9.0 and targeted for implementation in FPGA (DE-2 of decrypts, weighing more in the second, more precisely the
Amendment) capable of supporting up to 33.216 logic elements. inverse of the MixColumns operation, so the implementation
The results of Table 2 can be obtained.
 proposal should be compared to the decryption, since the union- [3] Daemen, J., and Rijmen, V. "The Rijndael Block Cipher,
inverse transformation does not compromise much run time. AES Proposal: Rijndael". Link in:
http://csrc.nist.gov/archive/aes/rijndael/Rijndael-
This implementation also enables the use of pipeline - proposed ammended.pdf.
future implementation. Could be a sharing of transformations
for up to three blocks of the state. Thus, the throughput obtained [4] Mathias, L. A. P. "Algoritmo AES - Relatório técnico da
for one block would be tripled, thereby increasing its speed of disciplina Redes I", UFRJ, Julho 2011. Link in:
execution. http://www.gta.ufrj.br/grad/05_2/aes/.
[5] Ordoñez, E. D. M., Pereira, F. D., Penteado, C. G., and
7. REFERENCES Pericine, R. de A., "Projeto, Desempenho e Aplicação de
[1] Daemen, J., and Rijmen, V. "AES submission document Sistemas Digitais em Circuitos Programáveis (FPGAs)".
on Rijndael - version 2", September 1999. Marília – SP: Bless, 2004.
[2] FIPS PUB 197, "Advanced Encryption Standard (AES)", [6] Milene, J. S., "Fields and Galois Theory - Version 4.22",
National Institute of Standards and technology, U.S. March 30, 2011. Link in:
Department of Commerce, November 2001. Link in: http://www.jmilne.org/math/CourseNotes/FT.pdf.
http://csrc.nist.gov/publications/fips/fips197/fips-197.pdf.

View publication stats