ACM Transactions on Modeling and Computer Simulation · 2007 · 42 citations · 14 references
ComputingComputer EngineeringComputer ScienceAlgorithmsCryptographyNumber TheoryGenerating FunctionsRandom Number GenerationGeneratorsInitial State EffectsInitializationLinear Recurrence RelationsPseudo-random Number GenerationPseudo-random Number GeneratorsPseudorandom GeneratorsRandom Number Generator Weaknesses
We demonstrate that a majority of modern random number generators, such as the newest version of rand.c, ranlux, and combined multiple recursive generators, have some manifest correlations in their outputs if the initial state is filled up using another linear recurrence with similar modulus. Among 58 available generators in the GNU scientific library, 40 show such defects. This is not because of the recursion, but because of carelessly chosen initialization schemes in the implementations. A good initialization scheme eliminates this phenomenon.
14
Makoto Matsumoto, Takuji Nishimura · ACM Transactions on Modeling and Computer Simulation · 1998 · 5.7K citations
Differential cryptanalysis of DES-like cryptosystems
Eli Biham, Adi Shamir · Journal of Cryptology · 1991 · 2.9K citations
An Object-Oriented Random-Number Package with Many Long Streams and Substreams
Pierre L’Ecuyer, Richard Simard, E. Jack Chen et al. · Operations Research · 2002 · 343 citations