Entropic analysis in financial forensics

Abstract

Digital analysis has become an accepted tool for computer forensic examination of suspicious financial transactions. A major component of digital analysis is Benford's Law which postulates that in accounting data not subject to substantial error or fraud, significant digits are not uniformly distributed and that smaller digits occur more often than larger digits as significant digits. In this paper, entropy-based mutual information is proposed to decide whether a dataset conforms to Benford distribution or not. If the dataset is found to deviate from Benford distribution, further analysis may be made into the authenticity of data entries.