Resilience of Error Correction Codes Against Harsh Electromagnetic Disturbances: Fault Elimination for Triplication-Based Error Correction Codes

Abstract

Modern safety-critical systems rely heavily on robust communication channels. Even though these communication channels can be protected by Error Detection and Correction Codes, vulnerabilities caused by False Negatives still exist. These False Negatives can be caused by harsh electromagnetic disruptions and are detriment to overall safety. This paper considers the construction and structure of Triplication-based Error Correction Codes to find the most EMI-resilient code. Each code is tested and simulated in terms of the occurrence rate of False Negatives under single-frequency disturbances. It is found that a code with inversion is significantly more robust to these disturbances. Furthermore, a systematic fault injection is also performed to search for vulnerabilities within the code itself. The systematic approach allows us to reverse-engineer the expected vulnerabilities to real-world disturbances.

Complementary Fault Models for Assessing the Effectiveness of Hamming Codes

“…Alongside repetitiveness, all-zero or all-one code words should be avoided in order to lower the rate of False Negatives. These conclusions were drawn in [5] and specific code alterations to lower (and possibly eliminate) the occurrences of False Negatives are given in [2].…”

“…When speaking in binary terms, injecting a bit-flip is done by performing the XOR operation between the bit value and a logical '1'. The XOR operation then transforms a binary '1' into a '0' and vise versa, as illustrated by (2).…”

“…This approach is easily extended to incorporate other EDCCs, as was done for a Triplication-based fault injection campaign in [2]. It should be noted that the duration of this fault injection campaign increases exponentially with the code length: the Hamming Code has 2 7 -1 possible corruptions, whereas the triplication has 2 12 -1 possible corruptions.…”

“…The EMI fault model was already extensively used and described in previous work [2][3][4][5]. This paper focuses on the complementarity of both models thus only a summary of the implementation and the results is provided here.…”

“…Both fault models under consideration within this paper have already been used in previous experiments and simulations. The systematic fault injection model was introduced and described in [2]. It was used to determine the strength and resiliency of different triplication methods.…”