Persistent Fault Injection In Fpga Via Bram Modification

The feasibility of persistent fault analysis relies on special faults which can persist in all the rounds of block ciphers. This prerequisite can be positioned as a good fit into the FPGA scenario, which however has not been carefully exploited ever before. In this paper, we propose the persistent fault attack on the block cipher AES-128 implemented in FPGA where a new type of persistent fault is induced with the technique of Block RAM (BRAM) modification. The details of persistent fault injection are elaborated, especially on how the target bits of AES in BRAM can be identified and how they can be altered. Our experimental results show that: with the proposed attack, a simple statistical analysis can extract the secret key of AES-128 with S-Box implemented in BRAMs and protected by the countermeasure of inversive decryption based dual modular redundancy.