Combining High-Level and Low-Level Approaches to Evaluate Software Implementations Robustness Against Multiple Fault Injection Attacks

Abstract

International audiencePhysical fault injections break security functionalities of algorithms by targeting their implementations. Software techniques strengthen such implementations to enhance their robustness against fault attacks. Exhaustively testing physical fault injections is time consuming and requires complex platforms. Simulation solutions are developed for this specific purpose. We chose two independent tools presented in 2014, the Laser Attack Robustness (Lazart) and the Embedded Fault Simulator (EFS) in order to evaluate software implementations against multiple fault injection attacks. Lazart and the EFS share the common goal that consists in detecting vulnerabilities in the code. However, they operate with different techniques , fault models and abstraction levels. This paper aims at exhibiting specific advantages of both approaches and proposes a combining scheme that emphasizes their complementary nature

The State of Fault Injection Vulnerability Detection

“…In [32] the authors propose combining the Lazart process with the Embedded Fault Simulator (EFS) [6]. This extends from the capabilities of Lazart alone by adding lower level fault injection analysis that is also embedded in the chip with the program.…”

An automated and scalable formal process for detecting fault injection vulnerabilities in binaries

“…However, to validate the experiments, a manual fault injection of setting the loop bound value to 10 was tested and found to be CA vulnerable with the process . 3 3 3 3 3 3 3 3 3 3 2 3 3 3 2 3 3 3 3 3 3 3 3 3 3 3 No Fault The works of Rivière et al, 12 Potet et al, 19 and Pattabiraman et al 33 all have similar proposals to the process here, albeit in more limited manners.…”

“…The strengths of this approach include the following. By operating directly upon the binary, fault injection vulnerabilities that cannot be detected in source languages or intermediate representations can be detected . Formal methods, model checking here, ensure the rigor of the analysis and so ensure that fault injection vulnerabilities that are detected are real and not false positives.…”

“…By operating directly upon the binary, fault injection vulnerabilities that cannot be detected in source languages or intermediate representations can be detected. 11,12 Formal methods, model checking here, ensure the rigor of the analysis and so ensure that fault injection vulnerabilities that are detected are real and not false positives. An automated process can be easily iterated over various fault injection models and approaches, and thus allows broad, or even complete, coverage of possible fault injection attacks.…”

“…This section discusses some key differences with respect to the process and approach used here. Several recent works have considered formal approaches and fault injection vulnerabilities 9,10,12,19,33. Both Christofi et al9 and Moro et al 10 use formal methods to prove the efficacy of countermeasures, in the work of Christofi et al 9 for one specific implementation and in the work of Moro et al 10 as applied to extremely small assembly code fragments.…”