Linear Feedback Shift Register Genetically Adjusted for Sequence Copying

International Journal of Computer Science & Information Technology (IJCSIT) Vol 15, No 2, April 2023
DOI:10.5121/ijcsit.2023.15203 27
LINEAR FEEDBACK SHIFT REGISTER GENETICALLY
ADJUSTED FOR SEQUENCE COPYING
Ugalde-Franco Juan Manuel, Martinez-Gonzalez Ricardo Francisco
and Mejia-Perez Juan Francisco
Electrics – Electronics Department, TecNM – Instituto Tecnologico de Veracruz,
Veracruz, Mexico
ABSTRACT
The present manuscript proposes a comparison between two types of shift registers; the first one, a
traditional one with a linear feedback (LFSR); on the other corner, a register designed one register with
genetically-controlled feedback loop (NLFGR). Being more traditional, the linear feedbacked register
works as reference to offer comparison metrics, meanwhile the NLFGR is our research focus, since our
proposal relies in its good features to replicate certain behaviours, with shorter parsing time than linear
feedback shift register.
KEYWORDS
Lineal Feedback Shift Register, Non-linear Feedback Genetically-controlled Register, Heuristic search,
Genetic Algorithms, Sequence reproduction.
1. INTRODUCTION
Generating pseudorandom sequences has application in many fields, since cryptography,
steganography, encoders and decoders for different communication protocols [1], [2], as a
security core used in satellite networks for cell phone communications [3], [2], even into Monte
Carlo test applications, developing and simulation environments [4].
A Pseudo Random Binary Sequence (PRBS) generator is an n-bit binary number generator circuit
that make a number per clock cycle for 2n-1
clock cycles [5]. Popularly a PRBS is implemented as
a linear feedback shift register [6], [7], [8]. Expressed in simple terms, a PRBS refers to its
function, while an LFSR is described as the implemented circuit. When an LFSR includes non-
linear elements such as genetic algorithms, it evolves from LFSR to become a non-linear
feedback genetic register (NLFGR), which has interesting properties in randomness fields for
generating sequences not deeply explored. However to exploit its full potential it has to be
simulated in software and implemented in hardware to evaluate its efficiency. A hardware
implementation in a FPGA fulfils translation from a high-level language as Python, that is a very
useful tool for testing draft designs
In this paper, we will discuss a comparison between LFSR and NLFGR features to replicate a
random sequences. Firstly, an LFSR and NLFGR is implemented using Python, and generating a
specific sequence, then we compare parsing times to estimate efficiency as generating engine

28
1.1. Linear feedback shift register (LFSR)
It is a shift register whose input bit is a linear function of its previous state; the most commonly
used linear function for handling individual bits is the exclusive-or (XOR). Therefore, an LFSR is
often a shift register whose input bit is driven by an XOR function of some bits of the same shift
register, creating a closed loop. Figure 1 shows a schematic that can help to illustrate prior
statement.
Figure 1. Example for 8-bits LFSR
In its operation, the initial value of the LFSR is called seed, and it must be different from '0';
since the operation of the register is deterministic, and its produced values are completely locked
to its current state. In the same way, because the register has a finite number of possible states, it
must eventually enter into a repeating cycle.
The bits in the LFSR that influence the input are called taps. An LFSR produces a 2n-1
sequence
of states within itself, except for the '0' combination.
1.2. Genetic Algorithms
This is an iterative search technique inspired by the principles of natural selection. The GA does
not seek to model biological evolution but they origin optimization strategies. The concept is
based on the generation of individuals by the reproduction of their parents, and so on. During the
course of evolution, genes with slow evolution were replaced by genes with better evolutionary
adaptability; therefore, highly efficient strategies are expected in the individuals.
Many problems have complex objective functions, and their optimization tends to end at local
minimums-maximums ends. The idea of the GA is to optimize, by finding the maximum or
minimum of an objective function, for such, the principles of natural selection are used on the
function parameters.
The components of a genetic algorithm are:
• A function that you want to optimize.
• A group of candidates for the solution.
• An evaluation function that measures how candidates optimize the function.
• Mating function.

29
Simplifying, the GA contains genes (initial population), genetic evaluation (fitness function),
reproduction of alleles or gene values (crossover or recombination), allele mutation (randomly
changing a bit, or by using a specific function), until satisfy the aptitude function or exceeding
certain number of iterations.
2. STATE OF ART
The pseudo-random number generation technique implemented by a linear feedback shift register
(LFSR), or an LFSR arrangement, for granting non-linear attributes [9], [10], [11] at the data
encryption moment, it gives an important impact in security area, since it improves vulnerabilities
in the systems, highly increasing the security levels [2], [12]. However, the LFSR, at being a
linear device, it remains susceptible to attacks.
Antagonistically, cryptanalysis has shown that if the noisy prefix of the output sequence is
known, it is possible to reconstruct the initial state of the LFSR by a generalized correlation
attack. This kind of attack is based on resolution of the restricted edition distance between
sequences, and such is determined by the initial state of shift registers and intercepted noisy
output sequence. Where FPGAs have been successfully applied for cryptanalytic purposes,
particularly in the exhaustive search of keys which is a highly parallelisable task [13].
One of the proposals is to make the systems more robust, the passwords or keys must be
analysed; for such activity, cryptanalysts require ways to automate the process so that
cryptographic systems can be more efficiently tested. Evolutionary algorithms provide one of
these features since they are able to search optimal global solutions very quickly. The cuckoo
search algorithm (Cuckoo Search, CS) has been effectively used in cryptanalysis of conventional
encrypted systems [14].
Another way to find satisfactory solutions to these problems is to use heuristic goals. That is an
algorithm designed to solve a wide range of optimization problems without having to deeply
adjust after each problem [15].
3. DEVELOPING THE USED MODEL
3.1. LFSR
Taking as a core an 8-bit register, with taps in bits 6, 4, 3 and 1 with right-side shifting, the LFSR
shown in Figure 2 is obtained.
Figure 2. LSFR with taps in bits 6, 4, 3 and 1

30
Which will be used as a model, to be implemented in high-level language Python, and then
implemented in an FPGA using VHDL language.
Its operating structure will be described in the flow diagram shown in Figure 3, but not before
explaining the variables used in diagram. N = Number, K = Iteration control, B = Search , NA =
Random number, cor = LFSR function.
To start operation, initialize K = 0, then request N that is necessary for program startingand
basically it can be one of two options; enter ‘0’ will look for a default value "10101010" or when
entering another number, the program will require an 8-bit binary number to be assigned to B,
then an 8-bit random number is generated, which will enter in a while loop that applies the
function cor to NA and adds 1 in each K cycle until variable B is equal to the variable NA or it
exceeds the 1000 iterations.
From the brief previous description, it should be noted that the function cor is essentially the
LFSR with the features of Figure 3. N is the element to find, and NA is the initial seed for the
LFSR operation.
Figure 3. Flow chart of LFSR behaviour
3.2. NLFGR
Taking as a core an 8-bit register with an evaluation function for the mating execution and
mutation function in each cycle, the whole process is repeated until reach stop criterion. The
processing loop is depicted in the block diagram in Figure 4.

31
Figure 4. Flow chart for the proposed genetic algorithm
Its operating structure will be described in the flow diagram shown in Figure 5, but not before
explaining the used variables of the diagram. N = Number, K = iteration control, B = Search, NA
= Random Number, and Fitness, Reproduction and Mutation = function NLFGR.
To star operation, initialize K=0 then request N that it is necessary for program starting, and it
also consists of two options. Entering ‘0’, operation will look for a default value "10101010";
meanwhile entering another 8-bit number, it will be assigned to B, and used for an 8-bit random
number generation, which will enter in a while loop, where the fitness function is applied to NA,
returning eight parameters indicating 1 when B=NA or 0 when they are different in each bit;
afterwards, the mating function uses the parameters delivered by the fitness function to indicate
evaluated population improvement, this to determine which pair of bits is more suitable to
mating. Such activity involves crossing bits (10-01), or determine to which bit will be applied
Mutation process; nevertheless this process needs to be non-recurrent; therefore, in each iteration,
when data enters the function, it must satisfy two precedents, that states which data bits receives
the NOT operation, and it must be in the Fitness parameters, and 0-to-7 random number must be
equal to the bit weight after being inverted and being in the Fitness parameter, on the contrary,
function will return an unchanged NA. After adding 1 to K in each cycle, until B equals NA or K
reach fifty iterations.
From the previous description, it is necessary to specify the optimization strategies that integrates
the described genetic algorithm, where genes are the randomly generated number (NA), after
introduce the searching parameters (B = N), the allele is each bit that integrates NA, the
evaluation function is implemented by Fitness, where it is reflected the key of the genetic
algorithm; in other words, the suitable points, the reproduction function is implemented by the
same name function, that is responsable for crosssing the bits for a new generation. The mutation
function is implemented by same name function that randomly adds parameters of Fitness, and
avoids stagnation in the production of generations, attending a tendency toward improvement.

32
Figure 5. Flow chart of NLFGR behaviour
4. RESULTS
From the LSFR and NLFGR algorithms, a series of fifty tests with the same word was realised to
reach "10101010" for measuring its operation performance and the following was determined:
• NLFGR consumes 10 times fewer iterations to find the result from a random seed.
• LFSR consumes 20 times more time to find the result from a random seed.
• The normal distribution of the two linear feedback shift registers show us the number of
iterations respectively to the standard deviation and its average result, verifying with its
graph the stability and operational performance under an observation period.
• The Table 1 shows a performance (seed-result) between experiment and iteration number
with regard to the two algorithms operating with a random seed.
Table 1. General results of the experiment
Average of Iterations and Computational Time to Find
Results
Element Iteration Time
LFSR 121.88 261.8 ms
NLFGR 19.9 10.7 ms
4.1. Statistical distribution LFSR
With reference to a normal distribution, the equation 1 is applied to the data obtained from the
realised experiments, the general behaviour of the device can be graphically determined.
f (x)=
1
√2πσ2
e
(x− μ)2
2σ2
(1)

33
Where:
µ = average
π= Pi constant
σ= typical deviation
σ2
= variance
e = Euler constant
x = data
We observe the normal distribution in Figure 6 with a standard deviation of 75.8 iterations, a
mean value of 121.88 iterations, a variance of 5745.64. The x-axis represents the data, meanwhile
the y-axis density value. After graphing the Gauss bell, we can determine:
• The number of iterations needed to find a result, in general, is from 101 to 144 iterations.
• The consumption of minimum iterations is not a characteristic of this device, because it depends
properly on its initial state.
Figure 6. Normal distribution of an LFSR
4.2. Statistical distribution NLFGR
We observe the normal distribution in Figure 7 with a standard deviation of 8.90 iterations, a
mean value of 10.9 iterations, a variance of 79.21. The x-axis represents data, and y-axis the
density value. After graphing, the Gauss bell you can determine:
• Few iterations are required for obtaining a satisfying result.
• The number of iterations needed to find a result, in general, is between 7 to 17.

34
Figure 7. Normal Distribution of an NLFGR
4.3. Graphical response of the experiment-number of iterations relation
The Figure 8 depicts relation between the experiment and number of iterations necessary to reach
control result, it highlights the reduced number of iterations necessary to maintain or success the
NLFGR result, with regard to the conventional LSFR.
Figure 8. Response experiment-iteration
4.4. Graph for the computational time
The relationship between experiment and the needed time to find the control result is depicted in
Figure 9. We can highlight the efficiency of NLFGR with respect to the conventional LFSR, and
its tendency to use a same computational time.

35
Figure 9. Comparison of LFSR and NLFGR computational times
4.5. VHDL Simulation
We recreate the Python coding in VHDL language, and it implements the same behaviour in a
FPGA developing board. In Table 2, we present output sequences for both, Python and VHDL,
codes using same seed. It gives congruent results that represents certainty about VHDL
implementation. The Python data was collected from the console where the program runs.
Table 2. Results of the VHDL simulation
Output sequences
Iteration Python VHDL
1 10100011 10100011
2 01010001 01010001
3 10101000 10101000
4 01010100 01010100
5 10101010 10101010
With the Python data, the next step is the VHDL simulation, which is presented in Figure 10. The
upper signal is the clock signal; meanwhile, the next signal is our interested one, NLFGR output.
Values are delivered each rising edge of clock signal, but for the last one, that occurs because
such data is our desired one, so the systems do not generate any other result.
Figure 10. Simulation of the sequence delivered from the NLFSR described in VHDL code

36
Having explained the previous process, it can be confirmed that both designs exhibit the same
behaviour.
5. CONCLUSION
It is determined that an NLFGR is more efficient than an LFSR to find a result, since its working
features are focused on optimization strategies but its initial state. It is highlighted that elements
obtained by experimentation show a number of iterations relative low, constant and consistent to
satisfactorily find values. Another remarkable point is the consistency in the output data parsing
time for the NLFGR algorithm, because it reaches the goal output around the same number of
iterations for each try.
Nevertheless their performance, NLFGRs have some limitations. They need prior adjustment to
tune, and works adequately. With no adjustment, the register simply does not converge. Another
limitation is the complexity, in fact it need lesser time to converge, but each parsing cycle
requires more operation.
Finally, we can express that following top-bottom design paradigm is reliable, since we started at
high-level language generator algorithm, for next describing the same algorithm in low-level
design as VHDL, and both succeed to gives correct data, giving confidence and reliability to our
proposal.
REFERENCES
[1] MAO, Yaobin; CAO, Liu; LIU, Wenbo. Design and FPGA implementation of a pseudo-random bit
sequence generator using spatiotemporal chaos. En 2006 International Conference on
Communications, Circuits and Systems. IEEE, 2006. p. 2114-2118.
[2] ZENG, Guang; DONG, Xiaodai; BORNEMANN, Jens. Reconfigurable feedback shift register based
stream cipher for wireless sensor networks. IEEE Wireless Communications Letters, 2013, vol. 2, no
5, p. 559-562.
[3] SHAH, Trishla; UPADHYAY, Darshana. Design analysis of an n-Bit LFSR-based generic stream
cipher and its implementation discussion on hardware and software platforms. En Proceedings of the
International Congress on Information and Communication Technology: ICICT 2015, Volume 2.
Springer Singapore, 2016. p. 607-621.
[4] JUSTIN, Remya; MATHEW, Binu K.; ABE, Susan. FPGA implementation of high quality random
number generator using LUT based shift registers. Procedia Technology, 2016, vol. 24, p. 1155-1162.
[5] BABITHA, P. K.; THUSHARA, T.; DECHAKKA, M. P. FPGA based N-bit LFSR to generate
random sequence number. International Journal of Engineering Research and General Science, 2015,
vol. 3, no 3, p. 6-10.
[6] ZULFIKAR, Zulfikar; AWAY, Yuwaldi; RAFIQA, Shahnaz Noor. FPGA-based design system for a
two-segment Fibonacci LFSR random number generator. Int. J. Electr. Comput. Eng.(IJECE), 2017,
vol. 7, p. 1882.
[7] FÚSTER-SABATER, Amparo; CARDELL, Sara D. Cryptographic Properties of Equivalent Ciphers.
Procedia Computer Science, 2016, vol. 80, p. 2236-2240.
[8] QU, Bo, et al. Differential power analysis of stream ciphers with LFSRs. Computers & Mathematics
with Applications, 2013, vol. 65, no 9, p. 1291-1299.
[9] MA, Zhen; QI, Wen-Feng; TIAN, Tian. On the decomposition of an NFSR into the cascade
connection of an NFSR into an LFSR. Journal of Complexity, 2013, vol. 29, no 2, p. 173-181.
[10] PEINADO, Alberto; FÚSTER-SABATER, Amparo. Generation of pseudorandom binary sequences
by means of linear feedback shift registers (LFSRs) with dynamic feedback. Mathematical and
Computer Modelling, 2013, vol. 57, no 11-12, p. 2596-2604.
[11] HU, Chunqiang; LIAO, Xiaofeng; CHENG, Xiuzhen. Verifiable multi-secret sharing based on LFSR
sequences. Theoretical Computer Science, 2012, vol. 445, p. 52-62.

37
[12] SHINY, M. I., et al. LFSR based secured scan design testability techniques. Procedia computer
science, 2017, vol. 115, p. 174-181.
[13] BOJANIĆ, Slobodan, et al. FPGA for pseudorandom generator cryptanalysis. Microprocessors and
Microsystems, 2006, vol. 30, no 2, p. 63-71.
[14] DIN, Maiya, et al. Applying Cuckoo Search for analysis of LFSR based cryptosystem. Perspectives
in Science, 2016, vol. 8, p. 435-439.
[15] BOUSSAÏD, Ilhem; LEPAGNOT, Julien; SIARRY, Patrick. A survey on optimization
metaheuristics. Information sciences, 2013, vol. 237, p. 82-117.

Linear Feedback Shift Register Genetically Adjusted for Sequence Copying

More Related Content

Similar to Linear Feedback Shift Register Genetically Adjusted for Sequence Copying (20)

More from AIRCC Publishing Corporation (20)

Recently uploaded (20)

Linear Feedback Shift Register Genetically Adjusted for Sequence Copying