Masking Signature Data Errors of Software-Based Control Flow Checking Techniques Employing Redundancy Signature

摘要

With technology scaling, transient faults are becoming an increasing threat to hardware reliability. Commodity systems must be made resilient to these in-field faults through very low-cost resiliency solutions. Up to 77% of the transient faults cause Control Flow Errors (CFEs). Software-based control-flow checking techniques have emerged as promising low-cost and effective solutions. The signature monitoring method is the foundation of most of these control flow checking techniques. Some CFEs cannot be detected by previous control flow checking techniques when transient fault hit the software signature. A technique, masking signature data errors of software-based control flow checking techniques employing redundancy signature (CFCRS), with the ability to mask these CFEs is proposed in this paper. In CFCRS, these errors can be detected and corrected by triple redundancy signature. The experimental results demonstrated that CFCRS is able to mask all 2,000 injected faults in software signatures; It is reasonable and feasible to apply this technique on the former software-based control flow checking techniques due to its perfect correction coverage of CFEs caused by incorrect-signature and low overheads.