Doctor of Philosophy
We propose a general framework to design a general class of random number generators suit- able for both computer simulation and computer security applications. It can include newly pro- posed generators SAFE (Secure And Fast Encryption) and ChaCha, a variant of Salsa, one of the four finalists of the eSTREAM ciphers. Two requirements for ciphers to be considered se- cure is that they must be unpredictable with a nice distributional property. Proposed SAFE-NET is a network of n nodes with external pseudo-random number generators as inputs nodes, several inner layers of nodes with a sequence of random variates through ARX (Addition, Rotation, XOR) transformations to diffuse the components of the initial state vector. After several rounds of transformations (with complex inner connections) are done, the output layer with n nodes are outputted via additional transformations. By utilizing random number generators with desirable empirical properties, SAFE-NET injects randomness into the keystream generation process and constantly updates the cipher’s state with external pseudo-random numbers during each iteration. Through the integration of shuffle tables and advanced output functions, extra layers of security are provided, making it harder for attackers to exploit weaknesses in the cipher. Empirical results demonstrate that SAFE-NET requires fewer operations than ChaCha while still producing a sequence of uniformly distributed random numbers.
Dissertation or thesis originally submitted to ProQuest
McCurdy, Jonathan, "SAFE-NET: Secure and Fast Encryption using Network of Pseudo-Random Number Generators" (2023). Electronic Theses and Dissertations. 3017.