International Journal of Computer Applications (0975 – 8887)

Volume 78 – No.16, September 2013

True Random Number Generator using Fish Tank Image

Rajat Katyal Ankit Mishra Adarsh Baluni

University of Pune University of Pune University of Pune
S.C.O.E. – Dept. of I.T. S.K.N.C.O.E. – Dept. of I.T. S.K.N.C.O.E. – Dept. of I.T.
Vadgaon (Bk), Pune-41, India Vadgaon (Bk), Pune-41, India Vadgaon (Bk), Pune-41, India

ABSTRACT methods as peripherals for the modern day PC’s making them
A Pseudo Random Number Generator (PRNG) uses a very difficult to implement.
deterministic system and an initial seed to generate random
numbers. In order for the output sequence to be truly random, We have used a simple fish-tank setup as the variable
a truly random input seed is used. Most True Random environment and capturing its images over time as the initial
Number Generators (TRNG), use noise in the form nuclear seed. A different image is used every time to generate a set of
decay, atmospheric noise, electrical noise or Brownian motion random numbers. The image data provides the initial seed
as their initial seed. which is extracted using a reduce algorithm and hashing. This
truly random seed is then fed to a pseudo random algorithm, a
In order to reduce the computational complexity, we use a linear congruential generator in our case, to give a set of
simple setup of a fish tank as the variable environment, random numbers.
capturing its images over time. The image data is then applied
to a reduction algorithm and hash function to generate the 2. NEED
initial seed. We propose a cost efficient method of extracting Random numbers are required in a lot of areas. Games,
the true seed from the image data and applying it to a pseudo statistical sampling, Monte Carlo method simulation,
random generator, a Linear Congruential Generator (LCG) in gambling and lottery, sort and hash algorithms are a few of
our case to give true random numbers. the fields that utilize the element of randomness.
General Terms The extensive utility of random numbers lies in the real -
PRNG, TRNG, LCG world security systems for exchange of classified data.
Cryptography uses random numbers to a great extent for
Keywords example while generating encryption key pairs and digital
signatures requiring a high degree of randomness [10].
True random number generator, Image data

1. INTRODUCTION 3. GENERATION
Random numbers are the numbers that are part of a sequence 3.1 Existing Methodologies
in which values are uniformly distributed over a definite set The Hardware random number generators use different
and any future value cannot be reliably predicted based on the sources of entropy to calculate the initial seed values. These
existing set of values. The sequence is produced by what is generally require complex hardware setups. Some of the
usually called as a random number generator. There are two commonly used methods are:
types of random number generators: pseudo - random number
generator and true random number generator. A pseudo -  Radioactive material decay time
random number generator uses a deterministic algorithm  Atmospheric Noise
which produces random numbers based on some initial value
known as seed. Most random number sources actually utilize  Electrical noise from resistors or semiconductors
a pseudo-random generator. Since the output is purely a
A similar approach to ours is to use a randomness extractor
function of the seed data, the actual entropy of the output can
such as cryptographic hash function against certain non-
never exceed the entropy of the seed [2].In certain real world
uniformly random source.
secure systems the quality of random numbers is of
paramount importance. For such applications it is not This technique had been used by the Lavarnd generator that
recommendable to use PRNGs as they use a deterministic took images of the patterns of floating point materials in a
algorithm. The output for such PRNGs are same for a lava lamp, extracting the seed and feeding it to a pseudo
particular seed, hence it is obvious that they cannot be secure random number generator.
if the seed is predictable in any which way.
3.2 Our Approach
TRNGs however use physical phenomena that are
We have used a fish tank setup having over 20 live fishes as
unpredictable in nature to generate true random numbers. The
our variable environment source. Clicking its images over
physical phenomena that are currently used are radioactive
time we get the random and unpredictable image data. The
decay, atmospheric noise, nuclear decay, Brownian motion,
image data is then applied to a reduce algorithm and hashing
clock drifts etc [9]. These results are then applied to
to generate a true and unpredictable seed.
cryptographic hash function for obtaining uniformly
distributed outputs. Unfortunately one does not find these

38
 International Journal of Computer Applications (0975 – 8887)
Volume 78 – No.16, September 2013

This seed is fed to a pseudo random algorithm similar to that

of a linear congruential algorithm in order to generate truly
35000
random numbers [14]. 30000
We have used the GIF image format having 640x480 pixels, 25000
in order to reduce the computational time and complexity.
Two image samples are shown below along with their true 20000
seed values.
15000 Series1
3.3 Samples 10000
5000
0
0 20 40 60

Fig 3: Sample set of 50 possitive random numbers

A small sample set of 50 random numbers produced using our
generator is shown in Fig 3. The sample has a range of 0 -
32768. The figure illustrates the randomness of the generated
numbers.

3.4 FLOWCHART

IMAGE DATA

Fig 1: Sample Image A

REDUCE ALGORITHM
AND HASHING

INITIAL SEED

PSEUDO RANDOM
GENERATOR

Fig 2: Image Sample B

RANDOM
In order to generate the different set of random numbers, we
NUMBERS
use a different image samples. For the shown image samples,
the generated true initial seeds were:
Fig 4: Flowchart of True Random Number Generator
Image A: 16338
Image B: 3453
4. TESTING
There are a wide variety of testing methods available that are
carried out to check the quality of generated random numbers.
These are typically applied to a PRNG for its quality check or
any new type of random number generator [11].
Since we utilize an already proven pseudo random algorithm,
a linear congruential generator, we did a quality test for our
chosen variables and the initial seed. The observed pattern
showed a high quality of randomness.

39
 International Journal of Computer Applications (0975 – 8887)
Volume 78 – No.16, September 2013

A single type of test cannot determine the quality of using it directly, thereby reduce the GIF image conversion
randomness. Also in a few tests, a vague pattern of result can overhead.
actually mean higher quality of randomness as it is basically a
factor of unpredictability. 2. Output range: 16bit

A Runs Test carried out on 80 numbers (below Table 1) yields The 16bit range (0-65536) of random numbers can be
the following result. Such a test alone cannot be a quality increased by using a bigger sized compiler and micro-
testing factor for randomness, but is one of the essential tests. processor, while maintaining the high degree of randomness.
Some other known testing methods are are frequency test, 7. ACKNOWLEDGMENTS
serial test, poker test and gap test.
Our thanks to P.Hellekalek and everybody that has
Table 1: Runs Test carried on a sample of 80 numbers contributed in generating good quality random numbers. Also
special thanks to our friend Tejaswi Bhutada.
7053 7716 1270 6327 3041
8. REFERENCES
5846 3780 3705 444 6035 [1] P. Hellekalek, Good random number generators are (not
4682 6603 5612 8623 893 so) easy to find, Mathematics and Computers in
4319 8631 4384 980 4195 Simulation 46 (1998) 485-505.
1618 5393 3641 7440 7430 [2] Benjamin Jun and Paul Kocher, P. 1999 The Intel
7954 2538 5232 4324 5356 Random Number Generator. Technical Report.
1919 320 8676 1102 2864 University of Maryland at College Park.
4830 5704 749 1773 3347
[3] N. K Pareek, V. Patidar, K. K Sud, A Rndom Bit
7545 8503 914 3770 1524 Generator Using Chaotic Maps, International Journal of
2219 4633 3774 7635 1939 Network Security, Vol. 10, no. 1, pp. 32-38, 2010.
63 2391 5476 1662 2766
[4] Tang, H.C., “An Analysis of Linear Congruential
2541 3267 7942 1667 8222 Random Number Generators when Multiplier
5249 1920 8396 1151 5693 Restrictions Exist,” European Journal of Operational
1683 6473 3452 270 6502 Research, Vol. 182, pp. 820828 (2007).
1246 4585 290 2133 8577 [5] Hedayatpour, S., Chuprat, Suriayati, Hash functions-
7928 2106 69 7192 3981 based random number generator with image data source,
Sample Set 80 Open Systems (ICOS), 2011 IEEE.
Number of Runs(R) 50
[6] Xuan Li, Guoji Zhang, Yuliang Liao, Chaos-based true
P- Value 0.02107 random number generator using image
Conclusion Moderate evidence
against Sequential [7] P. Murali and R.Palraj, True Random Number Generator
Randomness Based on Image for Key Exchange program.
[8] Xuan Li; Guoji Zhang; Yuliang Liao, "Chaos-based true
random number generator using image," Computer
5. RESULT AND CONCLUSION Science and Service System (CSSS), 2011 International
The random number generator is truly random as the seed Conference on , vol., no., pp.2145,2147, 27-29 June 2011
value used is taken from an unpredictable source. Hence it doi: 10.1109/CSSS.2011.5974933
completely avoids the possibility of determinism which is a
[9] Matsumoto, M. and T. Nishimura. "Mersenne Twister: A
risk factor in pseudo random number generators. 623-Dimensionally Equidistributed Uniform
The generator yields truly random numbers up to 16 bits or Pseudorandom Number Generator." ACM Transactions
65536 with a high degree of randomness. The random number on Modeling and Computer Simulation 8, no. 1 (1998):
generation can be divided mainly into two phases: Image to 3|30.
seed conversion and Pseudo random generation. [10] Volos, C.K.; Kyprianidis, I. M.; Stouboulos, I. N.,
It is also much easier to implement as compared to other "Image encryption process based on a chaotic True
hardware random number generators which use radioactive Random Bit Generator," Digital Signal Processing, 2009
decay, atmospheric noise, electrical noise, etc. and require a 16th International Conference on , vol., no., pp.1,4, 5-7
much complex hardware setup to generate randomness. The July 2009 doi: 10.1109/ICDSP.2009.5201107
TRNG using fish tank image data is a cost efficient alternative [11] Gentle, J. E. Random Number Generation and Monte
to generate truly random numbers. Carlo Methods, (2nd Ed.) SpringerVerlag, 2003.
6. LIMITATIONS AND FUTURE SCOPE [12] Geman, S. and D. Geman. "Stochastic Relaxation, Gibbs
1. Image processing time period Distributions, and the Bayesian

Although the pseudo random processing time is comparable to [13] Restoration of Images." IEEE Transactions on Pattern
any good PRNG but it involves an additional overhead of Analysis and Machine Intelligence 6, no. 6 (1984):
image processing time which makes it slightly slower. 721|741

The Future Scope of the True Random Number Generator is [14] Junod, P. "Cryptographic Secure Pseudo-Random Bits
to reduce the overall processing time. This can be done by Generation: The Blum|Blum|Shub Generator." August
using a camera to generate the hexadecimal image data and 1999.http://crypto.junod.info/bbs.pdf

IJCATM : www.ijcaonline.org 40