A Goal-Oriented Fault Tolerance Software Enhanced by Critical Information Code Protection

• Main Authors: Alban, Ahmad Qadeib, Jaoua, Ali
• Format: Conference Proceeding
• Language: English
• Subjects: attested process; Convergence; Critical Information; Fault tolerance; Fault tolerant systems; Fault-Tolerant Software; Hardware; SGX Technology; Software; Testing; Transient analysis
• ISSN: 2161-5330
• DOI: 10.1109/AICCSA50499.2020.9316510

Currently, during runtime, programs are mostly uncontrollable objects. They are very vulnerable to either transient or permanent, external, or internal contaminations of the program state. Here one finds a new method that combines a theoretical approach for goal-oriented software fault tolerance with SGX technology. Our approach consists to protect program critical information, by using enclaves in SGX security technology, in addition to checking regularly their preservation. Each module in the software is split into two parts: trusted codes protecting any operation on critical information, and an untrusted code for the remaining operations. The program contains two kinds of recovery routines: first, recovering against endless loops, and second, a goal-oriented recovery by repeating each module call until consensus (i.e. two consecutive executions of the module give the same output). Applied to several programs, it enabled different transient faults recovery successfully with minor time complexity overhead. The method proposes a new fault-tolerant program structure. Designers should use such kind of approach in a systematic way to avoid software failure caused by several types of transient faults. By protecting the code that preserves critical information, we reinforce the initial fault tolerance of the programs. This solution is convenient for critical applications requiring high security.