The concealed information test (CIT) is widely used for detecting deception in criminal investigations, primarily leveraging the P300 component of electroencephalogram (EEG) signals. However, the traditional bootstrapped amplitude difference (BAD) method struggles to accurately differentiate deceptive individuals from innocent ones when irrelevant stimuli carry familiarity or inherent meaning, thus limiting its practical applicability in real-world investigations. This study aimed to enhance the deception detection capability of the P300-based CIT, particularly under conditions of low stimulus heterogeneity. To closely simulate realistic investigative scenarios, we designed a realistic mock-crime setup in which participants were familiarized with all CIT stimuli except the target stimulus. EEG data acquired during CIT sessions were analyzed using the BAD method, machine learning algorithms, and deep learning (DL) methods (ShallowNet and EEGNet). Among these techniques, EEGNet demonstrated the highest deception detection accuracy at 86.67%, when employing our proposed data augmentation approach. Overall, DL methods could significantly improve the accuracy of deception detection under challenging conditions of low stimulus heterogeneity by effectively capturing subtle cognitive responses not accessible through handcrafted features. To the best of our knowledge, this is the first study that employed DL approaches for subject-independent deception classification using the CIT paradigm.