Secure Tensor Decomposition Using Fully Homomorphic Encryption Scheme

Secure Tensor Decomposition Using Fully Homomorphic
Encryption Scheme:
Abstract:
As the rapidly growing volume of data are beyond the capabilities of
many computing infrastructures, to securely process them on cloud has
become a preferred solution which can both utilize the powerful
capabilities provided by cloud and protect data privacy. This paper
presents an approach to securely decompose a tensor, a mathematical
model widely used in data-intensive applications, to a core tensor
multiplied with a certain number of truncated orthogonal bases. The
unstructured, semi-structured, and structured data are represented as
low-order sub-tensors which are then encrypted using the fully
homomorphic encryption scheme. A unified high-order cipher tensor
model is constructed by collecting all the cipher sub-tensors and
embedding them to a base tensor space. The cipher tensor is
decomposed through a proposed secure algorithm, in which the square
root operations are eliminated during the Lanczos procedure. Theoretical
analyses of the algorithm in terms of time complexity, memory usage,
decomposition accuracy, and data security are provided. Experimental
results demonstrate that the approach can securely decompose a tensor.
With the advancement of fully homomorphic encryption scheme, it can
be expected that the secure tensor decomposition approach has the
potential to be applied on cloud for privacy-preserving data processing.
Existing System:
Cloud Computing is a concept that has many computers
interconnected through a real time network like internet. Cloud
computing is a set of IT services that are provided to a customer over a
network on a leased basis and with the ability to scale up or down their
service requirements. Cloud computing provides facility to share
distributed resources and services that belong to different organizations
or sites. In private cloud system, information is shared among the
persons who are in that cloud. In this paper we have proposed new
authentication system for cloud computing platform. The Proposed
technique based on encrypted one time password (EOTP). In the
proposed technique one time password is encrypted by public key of
user to obtain Encrypted one time password (EOTP). In the proposed
system third party is not required.
Proposed System:
The size of data in many fields is rapidly increasing towards
Terabyte level or even Petabyte level, as well as the data structures are
becoming more varied. The large scale heterogeneous data have posed
great challenges on current computing infrastructures, and new
approaches are in urgent need to address them. A cloud infrastructure is
the collection of hardware and software which can provide capabilities
to the consumers on a pay-per-use or chargeper-use basis. Many
heterogenous data are modeled as tensors .a type of high dimension
matrix widely used in many applications. Tensor decomposition is a
powerful tool to extract valuable information from large scale raw data.
This paper presents a new computing approach which can securely
decompose the tensor model generated from large scale heterogeneous
data
 We present a holistic framework to address the problem of
secure tensor decomposition on cloud.
 The framework not only allows us to utilize the powerful
computational capabilities of the cloud, but also ensures data
security during the process of tensor decomposition.
 We introduce a Unified Cipher Tensor (UCT) model for
heterogeneous data representation. The detailed procedures of
how to encrypt the low-order sub-tensors constructed from
heterogeneous data as cipher counterparts using the fully
encryption scheme, as well as how to embed them to a base
tensor space to generate a unified cipher tensor model are
illustrated in this paper.
We propose to employ the Lanczos method to decompose the
generated cipher tensor model to a core tensor and a certain number of
truncated orthogonal bases. A secure tensor decomposition algorithm is
designed in which the nonhomomorphic square root operations are
removed during the Lanczos procedure. Theoretical analyses of the
algorithm in terms of time complexity, memory usage, decomposition
accuracy, and data security are provided. A tensor model is used to
illustrate the linear relations between the scalars, vectors, and other
tensors. Tensor is a generalization of a matrix model, which is usually
called multidimensional array. This on tensor decomposition, fully
homomorphic encryption scheme, and Lanczos method. Parallel
implementation of Lanczos algorithms are studied to improve the
efficiency. Many studies on tensor decomposition, fully homomorphic
encryption, and Lanczos method have been performed over the past few
decades. However, all the investigations mentioned above are
concentrated on special topics, no systematic research has been
dedicated to secure tensor decomposition of large scale heterogeneous
data.
Architecture Diagram:
Fig. Framework overview of the secure tensor decomposition approach.

Module Description
The modules are:
1.Data Representation, Encryption and Submission
2.Construction of Cipher Tensor
3.Secure Tensor Decomposition
4.Obtain the Plain Core Tensor and Bases
Data Representation, Encryption and Submission:
The heterogeneous data collected in the clients are represented as low-
order sub-tensors using the method proposed in our previous work
.Thenthe sub-tensors are encrypted using the fully homomorphic
encryption scheme and the generated cipher results are submitted to the
cloud for unification and decomposition. In Fig. the unstructured video
data V D, semi-structured XML document XD, and structured database
DB are transformed to cipher low-order sub-tensors T E V D, T E XD, T
E DB, respectively.
Construction of Cipher Tensor:
The generated sub-tensors T E V D, T E XD, and T E DB are then
embedded to a base tensor model Tbase ∈ RItim×Ispa×Iclt to generate a
unified cipher tensor model T E using the tensor extension operation T E
= Tbase ⇀ ×T E V D ⇀ ×T E XD ⇀ ×T E DB. The three orders Itim,
Ispa, Iclt of the base tensor model denote the time, space and client
characteristics.
Secure Tensor Decomposition:
After unfolding the unified cipher tensor T E to matrices T E (1), . . . , T
E (N) , where N is the number of orders of tensor T E, the
symmetrization transformation is performed on each tensor unfolding to
generate the symmetric matrix sym(T E (i) ) = T E (i) (T E (i) ) T, 1 ≤ i ≤
N. The eigen vectors of the symmetric matrix sym(T E (i) ) are
corresponding to the left singular vectors of matrix T E (i) . The Lanczos
method is employed to perform the eigen value decomposition, namely,
sym(T E (i) ) = U E i Λ E(U E i ) T. The cipher core tensor S E can be
computed by applying Eq. (1) to the truncated bases U E 1 , . . . , U E N
and the unified cipher tensor T E.
Obtain the Plain Core Tensor and Bases:
By decrypting the cipher core tensor and cipher truncated bases
generated in Step 3, the plain core tensor S and plain truncated
orthogonal bases U1, . . . , UN can be computed. As the homomorphism
are supported during the secure tensor decomposition, the generated
results are correct and are identical to that directly computed using the
plain data. This paper focuses on Step 2 and Step 3, which correspond to
the secure representation function fr and secure tensor decomposition
function fd.
Algorithm:1 Cipher Tensor Representation:
T E = fr(Du, Dsemi, Ds)
Input: The unstructured data Du, semi-structured data Dsemi, and
structured data Ds. Output: The unified cipher tensor model T E. 1:
Represent the local heterogeneous data as low-order sub-tensors, and
encrypt them to cipher low-order sub-tensors on clients. 2: Upload the
generated cipher sub-tensors to cloud. 3: Embed all the cipher sub-
tensors to the base tensor model Tbase ∈ RItim×Ispa×Iclt, and obtain the
unified cipher tensor model T E. 4: Unfold the cipher tensor to matrices
and generate the symmetric matrices for decomposition.
Algorithm: 2 Secure Tensor Decomposition on Cloud and

Client.
Input: The reduced cipher tensor T E. Output: The cipher core tensor S E
and cipher truncated orthogonal bases U E 1 , . . . , U E N . 1: Unfold the
cipher tensor to matrices and obtain the corresponding symmetric
matrices T E (i) (T E (i) ) T. 2: for each matrix T E (i) (T E (i) ) T, 1 ≤ i
≤ N, do 3: Initialize the parameters by setting j = 1, wj = random integer
vector, δj = w T j wj , β1 = 1, vj = T E (i) (T E (i) ) Twj . 4: while δj =
̸ 0
do 5: Compute w T j T E (i) (T E (i) ) Twj and obtain the parameter αj
by receiving the division result computed on the clients. 6: Compute
vector wj+1 = vj − αjwj . 7: Increase j by 1, namely, j = j + 1. 8: Replace
δj with w T j wj . 9: Send δj and δj−1 to the clients and receive the
division result βj . 10: Compute vector vj = T E (i) (T E (i) ) Twj −
βjwj−1. 11: end while 12: Construct the tridiagonal matrix L using the
generated elements αj , βj , and compute the eigen values and eigen
vectors on clients. 13: Compute the left singular vector matrices and
obtain the truncated orthogonal bases. 14: Obtain the cipher core tensor
S E using the equation S E = T E×1(U E 1 ) T×2(U E 2 ) T . . . ×N (U E
N ) T 15: Return tensor S E and the bases U E 1 , . . . , U E N . 16: end
for.
System Requirements
H/W System Configuration:-
Processor - Pentium –III

Speed - 1.1 Ghz
RAM - 256 MB(min)
Hard Disk - 20 GB
Key Board - Standard Windows Keyboard
Mouse - Two or Three Button Mouse
Monitor - SVGA
S/W System Configuration
 Operating System :Windows95/98/2000/XP /7
 Application Server : Tomcat5.0/6.X /8.X
 Front End : HTML, Java, Jsp
 Scripts : JavaScript,jquery,ajax
 Server side Script : Java Server Pages.

 Database Connectivity : Mysql.
CONCLUSION:
Aiming to propose an efficient approach that can securely
process large scale heterogeneous data, this paper formalizes the secure
tensor decomposition problem, and proposes a holistic solution
framework to address it. A unified cipher tensor model is presented to
integrate all the encrypted low-order sub-tensors as a unified model.
Concise examples are provided for illustrating the process of cipher
tensor construction and unfolding. A Lanczos-based secure tensor
decomposition algorithm is introduced, in which the non-homomorphic
square root operations in Lanczos procedure are removed. Theoretical
analyses in terms of time complexity, memory usage, decomposition
accuracy, and data security are provided. Some very preliminary
experiments are carried out to evaluate the performance of the presented
methods. The results support that the proposed approach is feasible and
can pave a way for secure data processing on cloud.

Secure Tensor Decomposition Using Fully Homomorphic Encryption Scheme

Uploaded by

Copyright:

Available Formats

Secure Tensor Decomposition Using Fully Homomorphic Encryption Scheme

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Secure Tensor Decomposition Using Fully Homomorphic Encryption Scheme

Uploaded by

Copyright:

Available Formats

Secure Tensor Decomposition Using Fully Homomorphic

Fig. Framework overview of the secure tensor decomposition approach.

Algorithm: 2 Secure Tensor Decomposition on Cloud and

H/W System Configuration:-

Processor - Pentium –III

RAM - 256 MB(min)

Key Board - Standard Windows Keyboard

Mouse - Two or Three Button Mouse

S/W System Configuration

 Operating System :Windows95/98/2000/XP /7

 Application Server : Tomcat5.0/6.X /8.X

 Front End : HTML, Java, Jsp

 Server side Script : Java Server Pages.

You might also like