Performance Analysis and Scope of Residue Number System in Digital Computing and Hardware Security

The Residue Number System (RNS) has become increasingly popular in digital signal processing and cryptography due to its innate parallelism and inherent non-linearity. RNS represents numbers as remainders when divided by chosen moduli and enables parallel processing, leading to efficient arithmetic operations. However, it requires computationally expensive conversions and careful selection of the moduli set to avoid precision loss and deal with overflow issues. In the field of hardware security, RNS is an attractive proposition because of its highly non-linear transfer function that makes it difficult for attackers to model or predict the behavior of the system. In this thesis, the scope of RNS in both efficient computing and hardware security has been analyzed thoroughly. The application of RNS in building a fast throughput Multiply and Accumulate (MAC) block through parallel processing has been demonstrated in a 65 nm CMOS process, where post-layout performance evaluations of the proposed RNS MAC demonstrate a 17% improvement in latency with an area-power consumption overhead of 12% when compared to the traditional binary MAC. Moreover, the promising aspect of the highly non-linear RNS in building a secure PUF has also been examined rigorously by carrying out ML attacks against different behavioral PUF circuit configurations.