BYTE-ORIENTED HASHING METHOD WITH ENHANCED DIFFUSION AND A SPECIALIZED MN PROCESSOR FOR PRIVATE CLOUD INFRASTRUCTURES

Abstract

This paper proposes a byte-oriented hashing method with an enhanced diffusion level, designed for streaming data processing in private (departmental) cloud infrastructures operating under strict latency and hardware resource constraints. The method is intended for detecting accidental and non-targeted data modifications, ensuring reliable deduplication and accurate block indexing in distributed storage environments. The proposed approach does not aim to achieve full cryptographic resistance against adaptive adversaries; instead, it provides a high level of diffusion tailored for engineering tasks related to data integrity verification and indexing. The method constructs two statistical arrays: K[v], representing byte occurrence frequencies, and S[v], capturing positional characteristics. An additional mixing stage g(K, S) is introduced prior to the reduction function C(·), which produces a fixed-length hash value. Based on this method, a specialized MN-processor with a PI–PC–PM–DO microarchitectural structure has been developed. The paper presents both functional and structural models of the processor, evaluates hardware resource utilization, and provides experimental performance results for CPU, FPGA, and ASIC implementations.

The main contributions of the study include: the formalization of a byte-oriented streaming method with enhanced diffusion; the design of an engineering-grade MN-processor microarchitecture for deterministic hardware implementation; experimental evaluation of throughput, latency, and cycles per hash; and positioning of the method as an engineering hash function for integrity verification and deduplication tasks in constrained cloud environments.