2014 Sixth International Conference on Computational Intelligence and Communication Networks

Detection of Data Leakage in Cloud Computing

Environment
Neeraj Kumar Vijay Katta Himanshu Mishra Hitendra Garg
Dept. of CSE Dept. of CSE Dept. of CSE Dept. of CSE
HCST,Mathura-INDIA HCST,Mathura-INDIA HCST,Mathura-INDIA HCST,Mathura-INDIA
[email protected] [email protected] [email protected] [email protected]

Abstract—In the recent years internet technologies has become

the backbone of any business organization. These organizations
use this facility to improve their efficiency by transferring data
from one location to another. But, there are number of threats in
transferring critical organizational data as any culprit employee
may public this data. This problem is known as data leakage
problem. In the proposed work, we are suggesting a model for
data leakage problem. In this model, our aim is to identify the
culprit who has leaked the critical organizational data.
Keywords—Bell-LaPadula model (BLP); Hash Function; AES;
Watermark; Message chaining

I. I NTRODUCTION
In the current business scenario, data leakage is a big Fig. 1: In the Bell-LaPadula model, each subject S has a
challenge as critical organizational data should be protected lattice of rights
from unauthorized access. Data leakage may be defined as the
accidental or intentional distribution of private organizational
data to the unauthorized entities. It is important to protect clearance/classification scheme is expressed in terms of a
the critical data from being misused by any unauthorized use. lattice[4][5] as shown in Figure-1.
Critical data include intellectual copy right information, patent
information, functional information etc.
In many organizations, this critical organizational data have AES algorithm and RSA algorithm shows good perfor-
been shared to many stakeholder outside the organizational mance among different symmetric and asymmetric encryption
premises. Therefore, it is difficult to identify the culprit, who technique[6] based on different performance factors such as
has leaked the data[1][2]. In the proposed work, our goal is to key value, computational speed and tenability. Various exper-
identify the guilty user when the organizational data have been imental factors were also analyzed based on text files used
leaked by some agent. In the proposed work, Bell-La Padula and experimental results proves that DES algorithm consumes
security model has been used which provide the analysis and least encryption time than AES but in terms of memory usage
design of secure computer systems. This model is called data AES uses least time than DES algorithm. In RSA encryption
confidentiality model. time is more and also memory usage is very high[7][8]. These
Bell-LaPadula model mainly focuses on data confidentiality techniques are useful for real-time encryption.
issues and provides controlled access to classified information. In other model, it has been shown a new comparative study
In contrast to the Biba-Integrity model which describes rule for between encrypting techniques based on nine factors like
the protection of data integrity[3]. In this formal model, the key length, cipher type, block size, developed, cryptanalysis
entities in an information system are divided into subjects and resistance, security, possibility key, possible ACSII printable
objects. The notion of a ”secure state” is defined, and it is character keys, time required to check all possible keys. Based
proven that each state transition preserves security by moving on these factors AES is better than DES and RSA[9]. It is
from one secure state to other secure state, thereby inductively also been discussed that DES is secret key based algorithm
proving that the system satisfies the security objectives of the suffers from key distribution and key agreement problems but
model. The Bell-LaPadula model is built on the concept of RSA consumes large amount of time to perform encryption
a state machine with a set of allowable states in a computer and decryption operation. It have been also observed that
system. A system state is defined to be secure if the only decryption of DES algorithm is better than other algorithms
permitted access modes of subjects to objects are in accordance in terms of throughput and power consumption[10].
with a security policy. To determine whether a specific access In the recent years, lots of changes happens in the field of
mode is allowed, the clearance level of a subject S is compared watermarking systems. Digital images are more popular than
to the classification level of the object O to determine if analog due to easy duplication and transmission on different
the subject is authorized for the specific access mode. The types of networks. Watermarking is used where authentica-

978-1-4799-6929-6/14 $31.00 © 2014 IEEE 804

803
DOI 10.1109/CICN.2014.172
10.1109/.172
 1) Security Model: Current access set Z=Triplets
(S=subject, O=object, A=attribute), and security level is
defined as pair of(C, S). C= classification i.e. Public,
Confidential, Secret, Top secret. S= category set-i.e. Military,
Air force, Defense, R and D. (c1,s1) dominates (c2,s2) iff
c1 ≥ c2 and S1 ⊃ S2. Level-1 dominate Level-2 because
they form lattice. The clearance/classification concepts are
expressed in terms of a lattice[4][14]. In Bell-LaPadula model
there are two basic strategies which defined the foundation
for secure access are-
2) Reading down (NRU): A subject S has only read access
Fig. 2: Information Flow to objects O whose security level L is below the subject’s
current clearance level. This prevents a subject from getting
access to information available in security levels higher than
its current clearance level.
tion or ownership is needed[11][12]. Watermarking is more
efficient tool in ownership claiming and fingerprinting of 3) Writing up (NWD): A subject S has only write access
digital data[11][12]. Watermark can be used to transmit secure to objects O whose security level L is higher than its current
message from one place to other place. clearance level. This prevents a subject to pass information
The computational cost and time complexity is the mea- from lower level to its current level.
sure problems with robust cryptographic algorithms. These
techniques use the concept of message authentication. These 4) Simple Security Property: (NRU- No Read Up) A
mentioned techniques ensure that any change in message can subject S at a given security level may not read an object O
be easily traced out (active attack) but it fails in case passive at a higher security level. For any (S, O, A)∈ Z if A includes
attack. Therefore, one single technique is required for both observation, then level(S) must dominate level (O) i.e. Any
message confidentiality and authentication. The main concern public user cannot read a top-secret document.
of the proposed work is to protect the secret information being 5) ”star”-Property: (NWD- No Write-Down) a subject S
transmitted. at a given security level must not write to any object O at a
This paper is structured as follows: In Section II proposed lower security level. If a subject S can observe O1 and modify
model is discussed. Section III contains Applications and O2, then level(O2) dominates level (O1) i.e. cannot copy top
Efficiency Measurement of the proposed model. Results Mea- secret files into secret files.
surements and conclusion have been discussed in Section IV
and Section V respectively. 6) Read-Only: The subject can only read the object.
7) Append-Only: The subject can only write to the object
II. P ROPOSED M ODEL but it cannot read.
In this proposed model we are providing the solution for 8) Execute-Only: The subject can execute the object but
critical Data Leakage problem. The proposed model has been can neither read nor write.
described in the following sections- 9) Read-Write: The subject has both read and writes per-
missions to the object.
A. Secured Environment Infrastructure The Read and Write access of bell-LaPadula model is ex-
plained in Figure-3 where every subject and object has their
We are using the concept of Bell-LaPadula Model for own clearance and classification level respectively by which
providing secured infrastructure, it is a state-machine model they are able to access the document.
and used to apply access control in different environment such
as-
Military security - Army, Air-force, Navy, NATO, NASA etc. 10)Tranquility Principle: The tranquility principle of the
Commercial security- Marketing Sales, Research and develop- Bell-LaPadula model states that the classification of a subject
ment, Human Resource department etc. or object does not change while it is being referenced[15].

B. Creating Watermark
In Bell-LaPadula model, information flow will be between the
high levels to low level it is shown in Figure-2. We define a In this model server will add an image logo to all the stored
state, if the system as a secured environment, and it follows documents and this image logo represents the organization.
some rule defined, as the allowed access mode of the any As we know that each intensity value in the image ranges
subject S, with the any object O is allowed, with respect to from 0 to (224 − 1), and for each of the three components
defined security policy. To find whether any specific access of color image as RED, GREEN and BLUE ranges from 0 to
mode will be allowed, the clearance of a subject S is compared (28 − 1). Each character, has their ASCII values ranges from 0
to the classification of the object O. i.e. S =(S1, S2, S3,..Sn), to (28 − 1). So, any text can be Inserted into the document by
O =(O1, O2, O3,..On) both S and O are combine and creating replacing the intensity value of pixel location, with the ASCII
up the security level used to determine if the subject S is value of character, which is needed to be hide and transmit
authorized for the specific access mode[13][5]. with the documents[16][17].

804
805
 TABLE I: Server Directory Table

SNo. Client-Id SHA-512(Hash) (m,n)

1. IDC1 M1 −−−−−
2. IDC2 M2 −−−−−
3. IDC3 M3 −−−−−
4. —– −−−−− −−−−−
5. —– −−−−− −−−−−
n. IDCn Mn −−−−−

1) Phase I: Calculation of all parameters:

• Calculate the cipher text, C, by using secret message
(IDC ), Encryption Key K and AES-128 encryption
Fig. 3: Read and Write access provided by the Bell-LaPadula Algorithm. It will be implemented using block cipher
model techniques[20].
• Calculate message authentication code, M , using
IDC , initial vector (IV ) and SHA-512 scheme. No-
tice that, M is generated by using secret message
(IDC ), not by using the cipher text C. This will
confuse the intruder, and will provide the extra level
of security [19].
• Calculate positioning pixels in the document D as:
◦ Row positioning pixel, m = I(1, 1) + 2
◦ Column positioning pixel, n = I(1, 2) + 2
2) Phase II: Placement of cipher C and authentication code
M into image:
• Replace the pixel value starting from (m, n) in the
Fig. 4: Client Server Registration original document, with the value of cipher text C.
Each block in cipher text will change exactly 16 pixel
bits in the original document D.
• Replace the last 64 pixel bits, with the authentication
So, first applying any cryptographic algorithm to text and then code M calculated, in reverse order. This will confuse
embedding the resulting ASCII to document. This process to intruders and provide the extra level of security.
ensures the necessary security. The key idea behind the imple- Finally the watermarked document WMD will be
mentation of this technique is to embed secret message into generated as the output for this process as shown in
the document with a computationally secured and time effec- Figure-5.
tive manner. Instead of using a high weighted cryptographic
algorithm like RSA, an effective light weighted algorithm C. Sending WMD to Client
like AES can be used with authentication scheme like SHA-
512[18][19]. The main focus of the proposed model is that only In this phase the created Watermarked document WMD
the registered user will be able to access the critical document will be send to the requested client along with server’s public
otherwise non-registered user has to first register itself with the key certificates(P KCserver ), which verify the integrity of the
server it is shown in Figure-4. This Watermarking technique genuine source of document.
discuss how and where to place the authentication code in the
critical document D. The server securely will maintain a server This Server will use the nonce (Cnonce ) in order to protect the
directory table for each registered client’s id which is shown man in middle attack. The send document will be encrypted
in TABLE-1. The input to the algorithm is original document with the public key of the client (P UC ) and contain the hashS
D, secret message (IDC ) and 128 bit key (K) used in AES- which is created by the server. The process of sending WMD
128 encryption scheme which produce the cipher text C as an to client is shown in Figure-6.
output.

Client will receive the send document and open with the help
In second phase of watermark embedding includes another of his private key P RC and verify the document, by creating
input IV , (initial value), which is 512 bits long, and used the hash of received WMD document. If the received hashS
for generating 512 bits long message authentication code M, is equal to the created hashC (ie. hashS = hashC ) then the
it embedded into the document D. The process of watermark document is not altered in between and document integrity is
embedding in document is describe in following phases- maintained[17].

805
806
 Fig. 7: The Process client id detection form WMD
Fig. 5: Process of Watermark Embedding

Fig. 6: Process of sending WMD to client

D. Detecting the Client id

In this scenario suppose clients has leaked the document by
some means and this document has value for that organization Fig. 8: Overall Working of the Proposed Model
in terms of money and reputation. In this step we are focusing
on detecting the client’s who has leaked the documents. The
watermark is extracted from the received watermarked docu-
ment(WMD) by applying reverse procedure of above proposed III. A PPLICATIONS AND E FFICIENCY M EASUREMENT
scheme. This is shown in Figure-7 and processed in two The proposed method is efficient to use with any size
phases- of documents. Here, we have used only client IDC with
the original document. Therefore, in this case if the hidden
1) Phase I: Point out the placement of cipher C and message is short, the changes made to the original document
authentication code M . is also short, resulting in less change in original document,
and hence intruder will not be able to analyze the secret.The
• Server will use the table where point (m, n) is stored.
proposed technique can be used with the different types of
Here we are using the same (m, n) value for all the
image such as Binary images, Gray scale images and Color
clients and it will be secret.
images etc. so it can be said that this technique is best suited
• Calculate the m and n value, and then point out the to transmit symmetric keys in secure manner. The technique is
starting position of C in watermark Document WMD. very economical, because it uses light cryptographic algorithm
AES-128 with SHA-512 to provide double security with half
• Find the authentication code M is in the last 64 pixels computational time[20][21]. If RSA is used then it leads to
in reverse order. two major problems[22]. First its key length is very high,
i.e., of 1024 bits, Second it will be more difficult to calculate
2) Phase II: Calculation and Verification of Secret Mes- exponential computations than simple computations needed in
sage IDC . AES which uses the key of length only 128 bits long.The
technique proposed in this paper uses the mixing concept to
• Secret message IDC is calculated by using K and generate message authentication code, and put it in reverse
AES-128 decryption algorithm. order to confuse the intruders[17][22].

• Server will verify this IDC by matching in the stored IV. R ESULTS M EASUREMENT
directory table and also verify the IDC as a correct
message, by calculating authentication code M’ from In this model we have chosen the AES model because it
IDC . If M and M’ both are equal, the extracted is faster in the process of encryption and decryption. To crack
message is correct. And the client will be identified the 128-bit AES key using a well-known brute force attack
who has leaked this data. The overall working of the it would take 1 billion years. AES is the successor of DES
proposed model is shown in the Figure-8. as standard symmetric encryption algorithm for US federal
organizations. AES accepts keys of 128, 192 or 256 bits (128

806
807
bits is already very unbreakable), uses 128-bit blocks (so no R EFERENCES
issue there), and is efficient in both software and hardware. It [1] Rohit Pol, Vishwajeet Thakur, Ruturaj Bhise, and A Kat. Data
was selected through an open competition involving hundreds leakage detection. International Journal of Engineering Research &
of cryptographers during several years. If we compare these Application, 2(3):404–410, 2012.
algorithms in terms of encryption and decryption we find that [2] Rupesh Mishra and DK Chitre. Data leakage and detection of guilty
time taken in AES encryption of message over different size agent. International Journal of Scientific & Engineering Research, 3(6),
of the message is less than DES and RSA. It can be verified in 2012.
Table-II and Table-III by comparing the encryption and decryp- [3] Kenneth J Biba. Integrity considerations for secure computer systems.
Technical report, DTIC Document, 1977.
tion time of AES with DES and RSA Algorithm[20][22].In
[4] David Elliott Bell. Bell–la padula model. Encyclopedia of Cryptography
the Table-III it shows the decryption time for different size and Security, pages 74–79, 2011.
of the messages AES takes less time over DES and RSA [5] Mukesh Singhal and Niranjan G Shivaratri. Advanced concepts in
Algorithm. In Table-IV where we analyze the different factors operating systems. McGraw-Hill, Inc., 1994.
[6] AL Jeeva, Dr V Palanisamy, and K Kanagaram. Comparative analysis
of performance efficiency and security measures of some encryption
TABLE II: Comparison of various packet sizes for DES, algorithms. International Journal of Engineering Research and Appli-
AES & RSA algorithm (Encryption Time) cations (IJERA) ISSN, pages 2248–9622, 2012.
[7] E Thambiraja, G Ramesh, and Dr R Umarani. A survey on various
Sno DES AES RSA Data Size most common encryption techniques. International journal of advanced
research in computer science and software engineering, 2(7):226–233,
1 3.0 1.6 7.3 153KB 2012.
2 3.2 1.7 10.0 118KB
3 2.0 1.7 8.5 196KB
[8] Bruce Schneier, John Kelsey, Doug Whiting, David Wagner, Chris Hall,
4 4.0 2.0 8.2 868KB and Niels Ferguson. Performance comparison of the aes submissions,
5 3.0 1.8 7.8 312KB 1999.
[9] Hamdan Alanazi, BB Zaidan, AA Zaidan, Hamid A Jalab, M Shabbir,
Yahya Al-Nabhani, et al. New comparative study between des, 3des
and aes within nine factors. arXiv preprint arXiv:1003.4085, 2010.
[10] Aman Kumar, Sudesh Jakhar, and Sunil Makkar. Distinction between
TABLE III: Comparison of various packet sizes for DES, secret key and public key cryptography with existing glitches. Indian
AES & RSA algorithm (Decryption Time) Journal of Education and Information Management, 1(9):392–395,
2012.
[11] Hitendra GARG and Suneeta AGARWAL. A secure image based
Sno DES AES RSA Data Size watermarking for 3d polygon mesh. SCIENCE AND TECHNOLOGY,
1 1.0 1.1 4.9 153KB 16(4):287–303, 2013.
2 1.2 1.2 5.0 118KB [12] Hitendra Garg and Suneeta Agrawal. Uniform repeated insertion of
3 1.4 1.24 5.9 196KB redundant watermark in 3d object. In Signal Processing and Integrated
4 1.8 1.2 5.1 868KB Networks (SPIN), 2014 International Conference on, pages 184–189.
5 1.6 1.3 5.1 312KB
IEEE, 2014.
[13] CISSP Susan Hansche, CISSP John Berti, and Chris Hare. Official
(ISC) 2 guide to the CISSP exam. CRC Press, 2003.
[14] D Elliott Bell and Leonard J La Padula. Secure computer system:
TABLE IV: Analysis of various factors Unified exposition and multics interpretation. Technical report, DTIC
Document, 1976.
[15] David Elliott Bell. Looking back at the bell-la padula model. In ACSAC,
Factor Analyzed DES AES RSA volume 5, pages 337–351, 2005.
Development Years 1977 2000 1978 [16] Frédéric Deguillaume, Sviatoslav V Voloshynovskiy, and Thierry Pun.
Key-Length (Bits) 56 128,192,256 ≤1024 Method for the estimation and recovering from general affine transforms
Nature of Algorithms Symmetric Symmetric Asymmetric in digital watermarking applications. In Electronic Imaging 2002, pages
Encryption/Decryption(Speed) Low High Medium 313–322. International Society for Optics and Photonics, 2002.
Nature of Security Attacks Inadequate Highly Secured Highly Secured
[17] Stallings William and William Stallings. Cryptography and Network
Security, 4/E. Pearson Education India, 2006.
which will shows the characteristics of the DES, AES and RSA [18] Achal Kumar and Vibhav Prakash Singh. Digital watermarking using
Algorithms[23]. color image processing using images for transmitting secret information.
[19] JJK RUANAIDH and T PUN. Rotation, scale and translation invari-
ant spread spectrum digital image watermarking. Signal processing,
V. C ONCLUSIONS AND F UTURE S COPES 66(3):303–317, 1998.
The proposed technique will provide better security against [20] NIST FIPS Pub. 197. Announcing the Advanced Encryption Standard
(AES), 2001.
data leakage problem. We can detect the data leaker in real time
[21] Máire McLoone and John V McCanny. Efficient single-chip imple-
by using this method. It also protect different types of active mentation of sha-384 and sha-512. In Field-Programmable Technology,
and passive attacks. The proposed technique is computationally 2002.(FPT). Proceedings. 2002 IEEE International Conference on,
cost effective in terms of time and space uses. Therefore, this pages 311–314. IEEE, 2002.
can be useful in distributed computing environment to protect [22] Jakob Jonsson and Burt Kaliski. Public-key cryptography standards
data from data leakage. The proposed technique is based on (pkcs)# 1: Rsa cryptography specifications version 2.1. 2003.
symmetric algorithm, therefore it is infeasible to extend this [23] B Padmavathi and S Ranjitha Kumari. A survey on performance
model for web environment where multiple number of users analysis of des, aes and rsa algorithm along with lsb substitution
technique. International Journal of Science and Research, 2(4), 2013.
frequently accessing the data object. We can also implement
this technique for asymmetric cryptography.

807
808