A Software-Based Error Detection Technique Using Encoded Signatures

Abstract

In this paper, a software-based control flow checking technique called SWTES (software-based error detection technique using encoded signatures) is presented and evaluated. This technique is processor independent and can be applied to any kind of processors and microcontrollers. To implement this technique, the program is partitioned to a set of blocks and the encoded signatures are assigned during the compile time. In the run-time, the signatures are compared with the expected ones by a monitoring routine. The proposed technique is experimentally evaluated on an ATMEL MCS51 microcontroller using software implemented fault injection (SWIFI). The results show that this technique detects about 90% of the injected errors. The memory overhead is about 135% on average, and the performance overhead varies between 11% and 191% depending on the workload used