Fibonacci LFSR vs. Galois LFSR: Which is More Vulnerable to Power Attacks?

Linear Feedback Shift Registers (LFSRs) with primitive connection polynomials as feedback functions are used as primary components of many stream ciphers and other cryptosystems. The motivation of our work is to demonstrate that though hardware implementation of Galois LFSR offers higher throughput than its Fibonacci counterpart, the former could be more susceptible to power analysis attacks. This gains more importance with the fact that both the LFSR configurations are theoretically equivalent. We propose a new attack strategy that deduces the initial state of a Galois LFSR by determining the LFSR output stream from the difference of power dissipation values in consecutive clock cycles. In addition, experimental results on power traces of both configurations implemented on SASEBO-GII board show that LFSR output stream retrieval from power dissipation values in Galois LFSR involve much less error in bit sequences compared to its Fibonacci counterpart.