International Journal of Computer Applications (0975 – 8887)

Volume 13– No.4, January 2011

Novel and Efficient Cellular Automata based

Symmetric Key Encryption Algorithm for Wireless
Sensor Networks
K.J. Jegadish Kumar K. Chenna Kesava S. Salivahanan
Assistant Professor Reddy Principal
SSN College of Principal SSN College of Engineering
Engineering Kalavakkam- Jyothismathi College of Kalavakkam-603110
603110 Chennai, India Engineering and Technology Chennai, India
Shamirpet-500078, India

ABSTRACT cells called its hood (usually including the cell itself) is
Our proposed L2D-CASKE:The lightweight 2- defined relative to the specified cell. An initial state (time
dimensional (2-D) cellular automata(CA) based symmetric t=0) is selected by assigning a state for each cell. A new
key encryption algorithm of a 128 bit length that accepts a generation is created (advancing t by 1), according to
key length of 128 bits have been designed as lightweight some fixed rule (generally, a mathematical function) that
encryption algorithm. As being light weight, it can be determines the new state of each cell in terms of the
easily applied on small devices such as wireless sensor current state of the cell and the states of the cells in its
motes, smart cards and other PDAs. The algorithm is an neighborhood. For example, the rule might be that the
iterated cipher consisting of repeated applications of a cell is "On" in the next generation if exactly two of
simple round transformation with different operations and the cells in its neighborhood are "On" in the current
different sequence in round. In this paper, the functionality generation; otherwise the cell is "Off" in the next
of our proposed algorithm was verified using MATLAB generation. Typically, the rule for updating the state of the
tools and hardware implementation of the algorithm using cells is the same for each cell and does not change over
Field Programmable Gate arrays (FPGAs) are presented. time, is applied to the whole grid simultaneously, though
For this purpose, an iterative loop design of our proposed exceptions are known.
algorithm is presented and implemented on FPGA. Beyond
its low cost performances, the proposed architecture is Encryption, by theory requires highly complex actions
fully flexible with any parameters and takes advantage of such as permuting, flipping and altering data in such a
generic VHDL coding. Our efficient implementation way that it is undecipherable and provides complex
achieves lower area, and considerably higher throughputs relationship with the original text and the keys. This
by synthesizing on the target platform Virtex-4 XC4VL25 relationship should be non-linear so that decryption
-10ff668 in Xilinx 9.1 process is as tough as possible. The encryption process
must be fast in time and cheap in terms of the
Keywords: Cellular automata, Symmetric Encryption, components involved. Cellular automata provides a
FPGAs, VHDL. basic structure for highly parallel and complex
operations upon which a basic encryption scheme can be
built. Cellular Automata based processor can be used to
1. INTRODUCTION compute and alter data with high degree of linearity and
Recent advancement of communication and computing complexity.
technologies introduces different types of portable devices
that populate in our day to day life. These devices have Cellular Automata can serve as a source of random
limited battery power, restricted storage and low numbers that are used for encrypting messages, running
computation power to bring the device in affordable cost simulations, and other purposes. According to Sarkar
and portable size. Information security is of primary 2000 [3], the configurations of a succession of
concern for all users irrespective of the computing device cellular automata generations can be used as a
being used. Among the different approaches for achieving random sequence. One-dimensional cellular automata are
information security, the present work concerns with normally used for this purpose. Sarkar cites several
symmetric key cryptography. Varieties of encryption specific applications that have been proposed for cellular
algorithms are available to encrypt the data. Execution of automata-generated random numbers, including private
the traditional encryption algorithms consumes time, space key cryptosystems, public key cryptosystems and hash
and energy. Moreover, side channel attacks are based on functions.Wolfram [4, 5] further reveals that the random
time and power that can be applied to the block ciphers number generator used for large integers in his
implemented on smart card technology [1]. Also protecting Mathematical computational software is based on the
implementation against these kinds of attacks is usually elementary cellular automata known as Rule 30. Wolfram
difficult and expensive. states that this particular cellular automata type is used as
a random number generator due to the fact that it has the
1.1 Cellular Automata interesting and useful property of being chaotic.
A Cellular Automaton (pl. cellular automata, abbreviated
as CA) is a regular grid of cells, each in one of a finite The paper is organized as section 2 describing studies of
number of states, such as "On" or "Off". The grid can be various symmetric key algorithms and its performances ,
of any finite number of dimensions. For each cell, a set of section 3 depicts our proposed algorithm in detail,Section

30
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

4 illustrates Matlab implementation and its result analysis key to produce the cipher text. The basic design goal of
to test the functionality of our algorithm. Then basic LCASE is to provide an efficient algorithm for both
cryptanalysis tests and its results were shown to prove our hardware and software implementations, as well as to
security level. Subsequently, hardware implementations meet the traditional security requirements. Following
of our algorithm are presented. Finally, conclusion is criteria are taken into consideration for designing the
expressed in section 5. proposed algorithm:

2. RELATED WORK - Resistant against conventional cryptanalysis attacks

Application of a cryptographically strong algorithm such - Simpler to implement in hardware with reduced cost
as AES-Rijndael [6] leads to significant transmission - Easily fit into resource constrained devices
delay, and requires many computations as well as large - Fast and code compactness on a wide range of platforms
storage capacity. It is infeasible to incorporate such - Resistant against timing analysis attack
strong cryptographic algorithms in resource
constrained devices. On the other hand, encryption The cipher has been designed primarily for fast and
algorithm ICEBERG [7] was proposed for easy implementation in both hardware and software
implementation with special emphasis on reconfiguring platforms. Parallel implementation at the bit-level can
the hardware. However, the software implementation is make the implementation faster. Therefore, the design
not suitable, i.e., not cost effective with respect to storage choices have been related to use CA-based components. A
requirement and/or speed. complete LCASE round comprises of first-half (FH)
round and last-half (LH) round. The encryption algorithm
Data dependent permutation (DDP)-based fast encryption is consisting of 'r' complete rounds while the last (r+1)
algorithms appear to be faster and efficient for high speed round is an half- round (FH-round) to make the cipher
networks. Recently, Cobra-H64 and Cobra- H128 were and its inverse to be similar in structure. The value of 'r' is
proposed in [8] which use switchable operations to chosen as 12/14/16 respectively for 128/192/256-bit user
prevent the weak keys identified against the earlier DDP- selected keys. Each round of encryption and decryption
based encryption techniques. These ciphers are specially process uses a key sub-block that is generated from user-
emphasized for high speed performance hardware selected key by a simple key schedule.
implementation but require more hardware resources. On
the other hand, Cobra-S128 [9] proposed for software Encryption is a one-to-one function for possible
implementation uses addition and subtraction modulo-232 decryption. Each encryption round uses one-dimensional
operations. These DDP-based techniques though (3-neighbourhood) 32-cell periodic boundary CA. Each
optimized for hardware or software implementation, round of encryption utilizes Reversible Cellular Automata
consume more resources (area, storage, computational (RCA), Bit Permutation (BP), Non-autonomous Cellular
power) making them unsuitable for implementation in Automata (NCA) and Reverse Substitution (RS).LCASE
resource constrained devices. A suitable alternative is to uses second order RCA based on CA rule 30, CA skewed
use techniques those are fast in terms of parallel rule 30, CA skewed rule 45 and NCA based on CA rule
operations and lightweight in terms of both computational 218. The choice of these rules has been opted from the
and storage requirements. pseudo-exhaustive simulation. It is observed that the
combination of these rules provide effective non-linearity.
The inherent parallelism property of Cellular Key mixing in the encryption process is not a simple
Automata (CA) has been exploited by several XOR of the key-bits but performs through an RCA or
researchers to design high speed encryption schemes NCA operation. RCA and NCA are the function of 3 key-
[10,11]. The cipher proposed in [11] cannot guarantee bits and one text- bit. This adds extra difficulty for an
security unless the size of the key is so high. Because the attacker to deduce the key from the given cipher.
same rule is applied repeatedly, the seed value, i.e., the
secret-key can be determined easily [12]. The cipher
proposed in [13] was broken in [14] due to the affine 3. OUR PROPOSED ALGORITHM
property of used CA in [15]. Recently, a reversible CA DESCRIPTION
(RCA) based encryption algorithm was proposed in [15] A complete L2D-CASKE round comprises of simple XOR
that satisfies the strict avalanche criteria, but trades-off , Block Permutation, 1D RCA Key Generation, 2D RCA
with additional communication overhead. In [16] a CA Margolus Neighborhood cell. The encryption algorithm is
based cryptosystem (CAC) is proposed, where non- consisting of r complete rounds. The values of r are chosen
linearity is achieved by intermixing affine CA with non- to be 12/14/16 respectively for 128/192/256- Bit user
affine transformations. selected keys. Each round of encryption/decryption
process uses a key sub-block that is generated from user-
2.1 LCASE Algorithm Review: selected key by a simple key schedule. Detail description of
Lightweight Cellular Automata-based Symmetric-key the proposed encryption algorithm and key schedule are
Encryption [2] is a simple one dimensional cellular discussed subsequently.
automata based symmetric-key encryption algorithm. The .
cipher uses a prevalent Rule 30 over the plain-text and the

31
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

FIGURE 1: Proposed Encryption Scheme

xi (t + 1) = (xi—1 (t) ⊕ (xi (t) ∨ xi+1 (t)) ⊕ xi (t − 1)).

3.1 1D RCA Key Generation:
Reversible CA (RCA)
A cellular automaton is said to be reversible, if the
previous configurations can be retrieved from the given
current configuration(s). Our proposed algorithm uses
second order reversible class CA (RCA) [17] in which,
the configuration of the ith state on clock cycle (t + 1) is
determined by the states of the n-neighborhood
configuration at clock cycle t and the self configuration at
(t − 1) clock cycle. In reverse, one can determine the
configuration at (t − 1) clock cycle from the
configurations at t and (t + 1) clock cycle. For example a
3-neighborhood second order RCA can be expressed as

xi (t + 1) =f (xi—1 (t), xi (t), xi+1 (t)) ⊕ xi (t − 1);

xi (t − 1) =f (xi—1 (t), xi (t), xi+1 (t)) ⊕ xi (t + 1).

Let, ξi and yi respectively denote for xi (t + 1) and xi (t

− 1). Depending on two initial configurations (Y, X) at FIGURE 2: Logic Cell of 4-cell Periodic Boundary RCA
[1]
time steps (t − 1) and t, the next configuration (ξ) is
evolved. Again, using two consecutive configurations
Initial configuration of RCA is setup with keyword kj l ,
(ξ, X ) the initial configuration Y can be deduced. We
where kj l denotes the jth key sub-block for lth round (1 ≤ l
denote these operations as
≤ r + 1). The previous configuration of that RCA is loaded
with the plaintext (intermediate ciphertext) sub-block X.
ξ = RCA (Y, X);Y = RCA (ξ,
The state of ith cell yi of the resultant configuration Y =
X).
RCA(X, K ) on CA rule 30 is expressed as in above
Equations.
Logic diagram of a second order periodic boundary 4-cell
RCA using elementary CA rule 30 is depicted in Figure 2.
The evolved configuration of such an RCA can be
expressed as

32
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

next-step cells, and the previous information could be

distributed to the next-step cells, conserving the total
amount of information. However, the possibility of this
happening does not seem very high. Here the Margolus
neighbourhood comes as shown in Figure 5(b).

Conventional CA (a)
MargolusNeighbourhood(b)

FIGURE 5: Conventional CA vs Margolus

FIGURE 3: A Simple Key Generation Algorithm Neighbourhood[18]

3.2 Block Permutation It uses as input the two cells of a block and returns as
The Block Permutation (BP) operation permutes each 2x2 outputs the new states of all two cells of the same block;
sub-block within a 4x4 block diagonally. The idea behind therefore, preservation of information can be assured. If
this permutation is to move the data at the boundaries at exactly the same partitioning scheme were to be used
the centre of the block and vice versa. This increases the repeatedly, information would be unable to propagate
rate of diffusion and makes differential cryptanalysis beyond each individual partition. To overcome this, the
difficult. Moreover, implementation of this permutation is Margolus neighborhood alternately groups odd and even
very simple and can be hard-wired simply by wire- cells every other step. Margolus neighborhoods are small
crossings. The block permutation operation is described enough that devising totalistic rules renders little benefit.
in Figure 4. The purpose of diffusion on a cipher is to On the other hand, Margolus was interested in reversible
spread out redundancy of the plain-text all over the cellular automata [19], for which rules encompassing
cipher-text, i.e., each bit of the plain-text needs to shifting, reflection, creation, annihilation and
contribute to as many cipher bits. Our proposed scheme complementation abound in such small neighborhoods; the
satisfies the diffusion requirement by the clever use of Billiard-Ball Machine was a result. One of the simplest
Block Permutation operation. partitioning schemes is the Margolus neighbourhood,
named after Norman Margolus and studied extensively
in a book he co-authored, Cellular Automata Machines
(CAM). This consists of dividing a grid of cells into
groups of four, as shown in Figure 6, to which the
automata rule is applied completely locally. If exactly the
same partitioning scheme were to be used repeatedly, then
FIGURE 4: Block Permutation Operation information would be unable to propagate beyond the
confines of any individual partition - and the dynamics of
3.3 2D RCA: Margolus Neighborhood the overall system would be sterile. The partitioning
Reversible Cellular Automata in Two Dimensions: As scheme is thus applied using grids that occupy different
we consider the plain-text to be of two-dimensions, we spatial co- ordinates on alternate time steps. The
make use of 2D Reversible Cellular Automata (RCA), one partitioning scheme for the alternating time steps is show
approach being Partitioned Cellular Automata (PCA) is in Figures 6 (a) and (b). The Margolus neighbourhood
another technique to generate invertible cellular partitions the space of the CA into disjoint boxes
automata. Margolus neighborhood, named after Norman consisting of 2 cells on a side. That partitioning then
Margolus, is a well-known partitioning technique. The idea changes from one clock tick to the next so that members of
is that conventional cellular automata are in general not one box become joined to members of an adjacent box on
invertible because their reverse step is not deterministic. alternate ticks. In Figure 6(a), the box boundaries are
As we can see in Figure 3, four possible past drawn so that all the even cells (added just for
configurations exist for either state of the current cell for clarification) and all the odd cells are grouped together,
the conventional Rule 30. Obviously, it is impossible to while in Figure 6(b), which represents the alternate half of
guess the past step with the current state of each cell. The the cycle, the cells have remained stationary while the
substantial reason is that conventional cellular automata boundary lines were shifted by one square in the diagonal
may lose some information while evolving. For instance, direction so that the odd and even cells share boxes. The
as shown in Figure 5(a), conventional one-dimensional other significant difference between the Margolus
cellular automata have three inputs, the cell itself and its neighbourhood and the conventional CA neighbourhood
two neighbours, but the output is only one cell. Having the is that the entire contents of the box are updated at once.
same number of states, it is very doubtful that the one All four cells change state based on the same rule which is
output cell can preserve all the information of the three written in the form known as D Rules: Billiard Ball
previous input cells. It is true that each cell affects three Machine rules.

33
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

the difference between the plain-text and the cipher-text

were obtained corresponding to each plain-text. Also, for
grayscale and colour images, Cipher Block Chaining
(CBC) was used, where the plain-texts are combined with
the cipher-text of the previous plain-text before applying
the cipher. Some of the results for plain-texts being a bit
image, a grayscale image and a colour image followed in
Figure 8 and 9

Odd Steps (a) Even Steps (b)

FIGURE 6: Margolus Neighbourhood Partitioning [20]

In the Margolus neighbourhood, the entire matrix is

divided into non-overlapping sub- blocks, each of 2x2
cells (bits), implying 2(2x2)=16 possible configurations
for every sub-block these 16 configurations are
assigned numbers ranging from 0 to 15 as shown in
Figure 7.

00: , 01: , 02: , 03: ,

FIGURE 8: Encryption and Decryption of

04: , 05: , 06: , 07: , Grayscale Image

08: , 09: , 10: , 11: ,

12: , 13: , 14: , 15:

FIGURE 7: Sub-Block Configurations [38]

4. IMPLEMENTATION OF
ALGORITHM AND RESULT
ANALYSIS
Our objective of designing and implementing a 2D CA-
based symmetric-key encryption algorithm has been
fulfilled. MATLAB was used as the development and
testing platform. The necessary criteria for a good cipher is
that the relationship between the plain-text, the key and the
cipher-text should be as involved and complex as possible, FIGURE 9: Encryption and Decryption of
making it almost impossible to decipher the plain-text Color Image
from the cipher-text without the exact key. The cipher-text 4.2 Basic Cryptanalysis Results:
on deciphering with the exact key should yield the plain-
text without any error. Various tests were performed on our In order to evaluate fitness, rules have to be subjected to
proposed algorithm to ensure that both the above criteria tests. Here is one of the interesting characteristics of CAs -
are satisfied. Our proposed algorithm was implemented it is impossible to assess each individual rule completely
straight forwardly by generic VHDL in Xilinx 9.1 tool to on its own. The whole idea is that individual
take advantage of the parallelism inherent in our design. components together lead to some meaningful
MATLAB is only a sequential programming platform behaviour and we do not know what each individual
whereas VHDL facilitates simultaneous parallel component is supposed to be doing. Behaviour of each
operations. The HDL modules were also synthesized and individual is too closely interwoven with behaviour of
the equivalent hardware circuitry of our algorithm was the cells in its neighbourhood. The only way to assign
extracted. Our efficient implementation achieves lower fitness to a rule is to run the entire CA on “many”
area, low power consumptions and considerably higher problems, each of which has a known solution, and see if
throughputs on the target platform SPARTAN-3, xc3s1400. the given cell reaches the correct “limiting” state in each
case. The ratio of the number of correct final states to the
4.1 Results in MATLAB: total number of tests gives a good approximation of the
As discussed above, the preliminary testing was done fitness of a cell. Of course, since the simulation has to end
using MATLAB. Plain-texts of size 128-bits and key in a reasonable time, both “many” and “limiting” have to
length of 128 bits were chosen and encrypted using the be given finite values that are sufficient and realistic. Some
proposed cipher. The cipher-text, the decrypted data and preliminary cryptanalysis was performed in MATLAB and
its results follow.

34
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

4.2.1 Basic Statistical Results:

The results of some basic statistical analysis such as the
distribution of ones and zeros, variation of ones in cipher-
text with respect to the plain-text, etc. are summarized
below.

4.2.1.1 Variation of Ones and Zeros - Balance

Property:
The number of ones and zeros (for randomly
generated key) in the various images showed in Figure
8 are tabulated below in Table 4.1

Table 4.1 Basic Statistical Results

4.2.1.2 Variation of Ones:
The following results (as shown in Figure 10) were
obtained by variants in the number of ones in the plain-
text between 0 and 128. It depicts the relationship
between the bit distributions of the plain-text and the
cipher-text for a 128-bit data. The graph shows minimal
variations in the characteristics of the cipher-text
throughout the large variations in the characteristics of
the corresponding plain-text. This is proof of the fact that
the statistical properties of the cipher-text are pseudo-
random and independent of the statistical properties of
the plain-text.

FIGURE 10: Variations of Ones in Encrypted Data

4.3 Hardware implementation FPGA-based computation engines appear to be very

Our iterative loop architecture for L2D-CASKE consists attractive for CA algorithms, which consist of a uniform
of round function and key schedule. Resource consuming structure composed of many Finite State Machines
blocks are the 1D RCA key generation module, XOR, and (FSMs), thus, matching the inherent design layout of the
2D Margolus Neighborhood module; the block FPGA hardware. As the whole CA lattice completely fits
permutations are implemented by swapping wires. into the FPGA space, the resulting computing system is

35
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

therefore simple. For CA computation, the whole lattice The scope for extension of this paper includes, the
is downloaded to the FPGA from the host machine and is use of cryptanalysis attacks such as differential
run for the required number of generations. Finally, the cryptanalysis attacks , linear cryptanalysis attack and its
resulting CA generation is uploaded back to the host variants, interpolation and algebraic attacks, related key
system. The LUT count for various block ciphers are attacks, equivalent key attacks, timing analysis attacks on
provided in the Table 4.2 Our proposed cipher uses 128- our proposed algorithm. Apart from the above, field
bit plaint-text block whereas the other ciphers use the optimization can also be done in terms of better schedules
standard 128-bit plain-text block. Our efficient such as non-autonomous cellular automata to mix the key
implementation achieves lower area expressed in terms of and plaintext. On the whole we believe our proposed
LUT and considerably higher throughputs on the target cellular automata based block cipher can be used
platform VIRTEX-4, XC4VL25 -10ff668. Finally, power effectively to construct a highly effective encryption
analysis are made on Xpower analyzer tool in Xilinx 9.1. scheme for resource constrained devices like, Wireless
sensor Networks, RFIDs and other low powered portable
The implementation results prove that our security devices.
processor occupying 642 slices in terms of area,
169.22mW of total power consumption and Throughput 6. REFERENCES
of 2136 Mbits/sec at the maximum work frequency of
334 MHz is compatible to any Main processor of a [1] P. Kocher, J.Jaffre and B.Jun, “Differential
sensor Node in Wireless sensor networks. Thus we have power analysis,” Crypto’99, LNCS 1666, pp.
enhance the performance of a sensor node by achieving 398-412, Springer-Verlag, 1999.
maximum level security and efficiency regarding to [2] S Tripathy and S Nandi,"LCASE: Lightweight
Area, Power and Throughputs. Cellular Automata-based Symmetric-key
Encryption”, International Journal of Network
Security, Vol.8,No.2 ,Mar. 2009.
[3] Palash Sarkar, “A Brief History of Cellular
automata”, Journal of ACM Computing Surveys
(CSUR), Volume 32 Issue 1, March 2000
[4] S. Wolfram, “Cryptography with Cellular
Hardware Automata,” Crypto ’85, LNCS 218, pp. 429-
Complexity in 432, Springer-Verlag, 1986.
terms of Total Power [5] S. Wolfram, “Random sequence generation by
Block Cipher
LUTCounts Consumption cellular automata,” Advances in Applied Maths,
,Throughput vol. 7,no. 2, pp. 123-169, 1986.
Mbits/sec [6] J. Daemen, and V. Rijmen, Specification for the
Advanced Encryption Standard (AES),
Springer-Verlag,2002.
582, NA NA [7] F. Standaert, G. Piret, G. Rouvroy, J.
LCASE[1]
Quisquater, and J. Legat, “ICEBERG : An
704, NA NA involutional cipher efficient for block
ICEBERG[7]
encryption in reconfigurable hard- ware,” FSE
3376,1563 ’04, LNCS 3017, pp. 279-299, Springer- Verlag,
NA
AES[6] Mbits/secs 2004.
[8] N. Sklavos, N. A. Moldovyan, and O.
642, 2136 Koufopavlou, “High speed networking: Design
169.22 mW and implementation of two new DDP-based
L2D-CASKE Mbits/secs
ciphers,” Mobile Networks and Applications-
MONET, vol. 25, no. 1-2, pp. 219-231, Springer-
Table 4.2 Comparison of Hardware Complexity in Verlag, 2005.
terms of 4-Input LUT Count and Power [9] N. A. Moldovyan, P. A. Moldovyan, and D.H.
Sum- merville, “On software implementation of
consumption
fast DDP- based ciphers,” Internatiol Journal of
Network Security, vol. 4, no. 1, pp. 81-89, 2007.
[10] S. Nandi, B. K. Kar, and P. P. Chaudhuri,
“Theory and application of cellular automata in
5. CONCLUSION cryptography,” IEEE Transaction on Computers,
We have proposed a light-weight symmetric-key vol. 43, no.12, pp. 1346-1357, 1994.
cryptosystem using Cellular Automata. This encryption [11] S. Wolfram “Cryptography with Cellular
scheme meets all the basic requirements meted out by the Automata,” Crypto ’85,LNCS 218,pp.429-
AES. Also, this being light-weight, it can be easily 432,Springer-Verlag, 1986.
applied on small devices such as sensor motes, smart [12] C.K.Koc, and A.M.Apohan, “Inversion of
cards, etc. The basic automata can be down scaled to cellular automata iteration,” IEE Proceedings
work even on a basic 8-bit processor. The major of Computer and Digital Technique, vol. 144,
advantage of our proposed scheme is the use of dynamic no. 5, pp. 279-284,1997.
key schedule which minimizes the memory requirement [13] S. Nandi, B. K. Kar, and P. P. Chaudhuri,
on the smaller devices. “Theory and application of cellular automata in
cryptogra- phy,” IEEE Transaction on

36
 International Journal of Computer Applications (0975 – 8887)
Volume 13– No.4, January 2011

Computers, vol. 43, no.12, pp. 1346-1357, 1994.

[14] S. Blackburn, S. Murphy, and K. Paterson, [18] "Cellular Automata Rules Lexicon - Margolus
“Comments on theory and application of cellular Neighborhood." Free Software of mirek
automata in cryptography,” IEEE Transactions on Wojtowicz.
Computers, vol. 46, no. 5, pp. 637-638, 1997. http://www.mirekw.com/ca/rullex_marg.html
[15] F. Seredynski, K. Pienkosz, and P. Bouvry, [19] Joaquin Cerda, Rafael Gadea, and Guillermo Paya,
“Reversible cellular automata based encryption,” “ Implementing a Margolus Neighborhood
NPC’04, LNCS 3222, pp. 411-418, Springer- Cellular automat on FPGA”, IWANN 2003,
Verlag, 2004. LNCS 2687, pp. 121-128, Springer-verlag Berlin
[16] S.Sen, C.Shaw, D.R.Chowdhuri, N. Ganguly, Heidelberg, 2003.
and P. Pal Chaudhuri, “Cellular automata based [20] "2D Reversible Cellular Automata."Lotus
cryptosystem (CAC)”, ACRI-2002, LNCS 2513, Artificial Life.
pp. 303-314, Springer-Verlag, 2002. <http://www.alife.co.uk/ca/bbm/2d/>
[17] T. Toffoli, and N. Margolus, “Invertible cellular
automata: A review,” Physica D, vol. 45, pp.
229-253, (reprinted with correction as of Oct.
2001).

37