Algebraic Fault Attacks

点击翻译按钮获取中文摘要

The protection against physical attacks plays a steadily growing, important role for the secure operation of mobile and embedded systems. The continued transition to cyberphysical systems, which are characterized by a high degree of connectivity and a tight coupling of (embedded) computers with the physical world and which often lack perimeter protection, increases the relevance of such attacks even further. The project “Algebraic Fault Attacks” investigates an essential class of physical attacks, namely fault-injection attacks, and the use of algebraic solving techniques as part of such attacks. In the first funding period of the project, solving methods based on a tight integration of border basis solvers with SAT solvers have been developed. Substantial advances have been achieved for both individual types of solving techniques, and a deeper understanding for their synergistic combination has been established. The attacks have been validated by measurements on an FPGA board, and all relevant data (algebraic models, circuit descriptions) have been made available on the project website.In the second funding period, the results achieved so far will be expanded. The attacks will be extended to further classes of crypto systems (stream ciphers, authenticated encryption, public-key and postquantum crypto systems), and the modeling via polynomial systems will be adjusted to the integrated solver. Solving methods will be extended by novel approaches: partial #SAT and approximate #SAT. Circuit descriptions will be incorporated into instance encoding, thereby further optimizing the combination of the solvers. Moreover, the interplay and mutual influence of fault attacks and other hardware-oriented threats, like passive side-channel analysis or reverse engineering, will be investigated. The consideration of a combination of multiple attack vectors is a natural prerequisite for the design of secure circuits, and it has received only cursory attention so far. Cross-level protection methods against fault attacks based on security-oriented nonlinear error-detecting codes will be developed. For the first time, the new constructions will be evaluated in realistic fault-injection scenarios. A bridge between information-theoretical and circuit-level models will be established, and the efficiency of countermeasures against multiple attacks vectors will be studied.