Identifying the Risk in Lie Detection for Assessing Guilty and Innocent Subjects for Healthcare Applications

Objective: This study aims to critically assess the implications and potential hazards of employing deep learning algorithms for the identification of innocence and guilt through lie detection methodologies within healthcare environments. Methods: The research centres on exploring Electroencephalography (EEG) as a significant technique for lie detection, particularly via the Concealed Information Test (CIT). CIT, a forensic approach, is deployed to identify hidden or deceptive information by monitoring physiological reactions to specific queries containing sensitive information pertinent to a case. This study highlights the importance of band-pass filters in lie detection, which enable the isolation of particular EEG signal frequencies that may reflect cognitive processes associated with deception. Findings: An exhaustive examination of various machine learning algorithms, including Support Vector Machines, Linear Support Vector Classification, Multilayer Perceptron, alongside deep learning frameworks like Long Short-Term Memory (LSTM) and Deep Neural Networks (DNN), was conducted, incorporating sophisticated feature extraction techniques. The investigation reveals critical considerations and potential risks linked with the application of these technologies for distinguishing between guilty and innocent subjects in healthcare applications, emphasizing the need for careful and ethical use of such advanced detection methods. Novelty: The outcomes of our study reveal that the Multilayer Perceptron (MLP) attained an impressive accuracy rate of 98.36%, while the Deep Neural Network (DNN) demonstrated a high accuracy of 98.28%.