Design space extension for secure implementation of block ciphers

Security has been identified as a critical dimension in the design of embedded systems for almost a decade. A well-recognised critical threat against the security of embedded systems is represented by ‘side-channel attacks (SCAs)’, which mandate the application of specially tailored countermeasures. These countermeasures are significantly demanding in terms of computation effort, and have traditionally been applied by hand. The recent introduction of a methodology to gauge the security margins provided by software cipher implementations, allows the integration of the automated application of countermeasures into platform-based system-level design methodologies. The authors introduce in the design space of block cipher implementations a new metric concerning the resistance against SCAs, provide a systematic method for the selection of the most appropriate cipher given the security and performance trade-offs, and point out the performance requirements for the random number generator. Moreover, they discuss the implications of the design space extension on system runtime adaptivity. The experimental evaluation demonstrates that a single cipher does not cover optimally a range of convenient operating points and that ciphers like a Serpent, which are considered slow in non-protected implementations, can outperform primitives like the Advanced Encryption Standard when implementations with equal security guarantees against SCAs are considered.