A comprehensive analysis of deception detection techniques leveraging machine learning

Abstract

Deception detection has recently drawn noticeable attention to an objective understanding of human behavior, which can be driven by various factors ranging from self-preservation to causing harm. Deceptive behavior can be categorized into various types based on its implications, as lies can range from being harmful to having serious consequences. Consequently, this field is paramount, particularly in critical areas such as the legal system, where accurate deception identification is essential. This paper presents a comprehensive analysis aimed at extracting and analyzing knowledge from existing literature on deception detection. Our approach aims to streamline deception detection methods based on machine learning and compares them to conventional non-machine learning approaches. Convolutional neural networks have demonstrated superior performance on real-life datasets compared to various modeling approaches. The findings reveal a common shortfall among existing studies, specifically the lack of consideration of cultural, linguistic, and gender influences on deception detection, as well as issues related to data scarcity and heterogeneity.