Cryptography Research, part of the Rambus group, is cooperating with French semiconductor security specialist INVIA in the next round of the ongoing battle against piracy and theft of service. The agreement will enable INVIA to develop products incorporating Differential Power Analysis (DPA) countermeasures for use by customers of its security-related semiconductor products, as well as by designers of ASICs and FPGAs that use its embedded software.
DPA is a sophisticated form of attack that involves monitoring fluctuations over microsecond time periods of the electrical power consumed by a target device. It then uses advanced statistical methods involving successive guesses of the digital information represented by the power patterns to deduce cryptographic keys and other secrets. The potential for such attacks has grown by development of better techniques, and through increasing computational power.
DPA countermeasures come in three broad categories. First is to reduce the variations in the electrical signals through techniques that leak less power. Physical shielding of the device could be one option. But, such techniques can never eliminate power leakage completely, so they cannot be guaranteed to protect against DPA attack by increasingly sensitive equipment, while making devices unacceptably expensive and big.
Therefore, a second approach was introduced, by adding noise to the system in order to increase the computational power and time needed to complete an attack, in the hope of making it unfeasible. But, techniques have evolved to overcome this defense, and so a third approach emerged, based on the assumption that neither of the first two defenses had yet been employed, and that the hardware to be protected does leak information.
Cryptographic Research seems to have focused particularly on the third approach, which can involve techniques that mash up successive keys and prevent attackers from accumulating the data from large numbers of sampling operations in order to perform the statistical calculations. Cryptographic Research calls this a leak-tolerant design, and argues that the advantage is that implementers can use masking techniques as required to meet specific targets determined in turn by the current capabilities of attackers. It really is another case of staying one step ahead of the attackers. The ongoing success of this particular method remains to be seen, but so far it has stood up to scrutiny.