Error Analysis of Forensic ENF Matching

This paper presents analytical results on the performances of several existing forensic electric network frequency (ENF) matching methods, i.e., benchmark minimum mean squared error (MMSE) based matching, error correction matching (ECM), and bitwise similarity matching (BSM). The use of a threshold value enables the ECM and BSM methods to tolerate certain amount of local matching errors, but performance improvement against the MMSE benchmark is not guaranteed at all times. By looking into the relationship between the local errors and the threshold value, we comprehensively investigate possible situations of matching inconsistency during ENF matching process, which has not been addressed in existing literature. We reveal that ECM generally yields the best matching results in the 7 possible matching situations discovered in this paper. However, the performance of BSM is inferior to the MMSE benchmark in several situations, although it enjoys significantly reduced computation time thanks to bitwise processing. This work indicates the potential to design improved matching methods.