Domain specific high-level synthesis of cryptographic workloads

As the world is getting closer to the paradigm of Internet of Things, ensuring security for all the phenomenally increasing information exchange is becoming more challenging. The VLSI designer for security solutions must judiciously juggle various resource constraints by weighing multiple possible microarchitectures in the entire design space before reaching an optimal solution for an application in mind. Additionally, the inherently dynamic nature of cryptography leverages newer cryptographic functions and attacks to frequently surface, shortening further the time to market. With traditional VLSI design cycle requiring significant man hours and sufficient design expertise, reaching in-time design optimality is hardly possible. Automation can expedite the VLSI design cycle. A rapid prototyping tool facilitating the automatic software and hardware implementation of cryptographic workloads is the major outcome of this thesis. The tool (called CRYKET) presents a language independent GUI based design capture, eliminating the steep learning curve often requried by traditional high-level synthesis tools. The cryptographic workloads were analyzed to identify a functionally complete set of constructive computation kernels as well as architectural sub-structures that are representative of a class of cryptography and their optimized hardware/software implementation. CRYKET allows fast design exploration by quick reconfiguration of design through a flexible API. A major technical challenge addressed by CRYKET is to seamlessly integrate these architectural substructures and operational functions into a working cipher model, without sacrificing the performance of the implementation, both of software and hardware platforms. Various symmetric key ciphers were successfully implemented and their performance was benchmarked to be on-par with their already-known hand-optimized HDL implementations. Other than the known microarchitectural optimizations, optimization strategies, which are too algorithm specific to be undertaken automatically by CRYKET are discussed, for example, flexible cryptographic processors supporting multiple cryptographic functions and emerging cryptanalytic schemes. Moreover, memory bank reorganization in the context of designing efficient cryptographic cores, and the scalability of cryptographic functions on modern multi-core GPUs is taken up. Development of a cryptanalytic hardware solution for attacking SHA-1 is also undertaken. These case studies can serve as a guide for future design exploration of efficient cryptographic implementations