Neurophysiological Approaches to Lie Detection: A Systematic Review

Background and Objectives: Lie detection is crucial in domains such as security, law enforcement, and clinical assessments. Traditional methods suffer from reliability issues and susceptibility to countermeasures. In recent years, electroencephalography (EEG) and particularly the Event-Related Potential (ERP) P300 component have gained prominence for identifying concealed information. This systematic review aims to evaluate recent studies (2017–2024) on EEG-based lie detection using ERP P300 responses, especially in relation to recognized and unrecognized face stimuli. The goal is to summarize commonly used EEG signal processing techniques, feature extraction methods, and classification algorithms, identifying those that yield the highest accuracy in lie detection tasks. Methods: This review followed PRISMA guidelines for systematic reviews. A comprehensive literature search was conducted using IEEE Xplore, Web of Science, Scopus, and Google Scholar, restricted to English-language articles from 2017 to 2024. Studies were included if they focused on EEG-based lie detection, utilized experimental protocols like Concealed Information Test (CIT), Guilty Knowledge Test (GKT), or Deceit Identification Test (DIT), and evaluated classification accuracy using ERP P300 components. Results: CIT with ERP P300 was the most frequently employed protocol. The most used preprocessing method was Bandpass Filtering (BPF), and the Discrete Wavelet Transform (DWT) emerged as the preferred feature extraction technique due to its suitability for non-stationary EEG signals. Among classification algorithms, Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Convolutional Neural Networks (CNN) were frequently utilized. These findings demonstrate the effectiveness of hybrid and deep learning-based models in enhancing classification performance. Conclusions: EEG-based lie detection, particularly using the ERP P300 response to face recognition tasks, shows promising accuracy and robustness compared to traditional polygraph methods. Combining advanced signal processing methods with machine learning and deep learning classifiers significantly improves performance. This review identifies the most effective methodologies and suggests that future research should focus on real-time applications, cross-individual generalization, and reducing system complexity to facilitate broader adoption.

Systematic review following PRISMA guidelines searching IEEE Xplore, Web of Science, Scopus, and Google Scholar for English-language studies from 2017–2024 that used EEG-based lie detection with CIT, GKT, or DIT protocols and evaluated P300 classification accuracy.

Background & Context

Lie detection has long been a critical challenge in forensic psychology, security screening, and law enforcement. Traditional polygraph methods suffer from reliability issues and susceptibility to countermeasures, with subjects able to manipulate physiological responses through training or medication. This systematic review addresses a paradigm shift in deception detection: the move from peripheral physiological measures to direct neurophysiological assessment of brain activity.

In recent years, electroencephalography (EEG) and particularly the Event-Related Potential (ERP) P300 component have gained prominence for identifying concealed information. The P300 is a positive-going brain wave occurring approximately 300 milliseconds after a meaningful or significant stimulus—including items an individual recognizes but is attempting to conceal. This neurophysiological marker has proven especially powerful in face recognition paradigms, where the brain's automatic response to familiar faces cannot be easily suppressed.

This 2025 review fills a critical gap by synthesizing recent advances in EEG-based deception detection from 2017–2024, a period marked by explosive growth in machine learning and deep learning applications. The review specifically focuses on identifying the most effective signal processing, feature extraction, and classification methods—providing evidence-based guidance for both researchers and practitioners seeking to implement next-generation lie detection systems.

Research Design & Methodology

This review followed PRISMA guidelines for systematic reviews, ensuring transparent and comprehensive literature identification. A comprehensive literature search was conducted using IEEE Xplore, Web of Science, Scopus, and Google Scholar, restricted to English-language articles from 2017 to 2024. The authors used targeted search queries tailored to each database's syntax, all variations of which combined "EEG" and "lie detection" terms.

Studies were included if they focused on EEG-based lie detection, utilized experimental protocols like Concealed Information Test (CIT), Guilty Knowledge Test (GKT), or Deceit Identification Test (DIT), and evaluated classification accuracy using ERP P300 components. The CIT paradigm presents subjects with both crime-relevant (probe) items and neutral (irrelevant) items, detecting recognition through differential P300 amplitudes.

The review systematically extracted and categorized information across multiple dimensions:

• Experimental protocols and stimulus types (particularly face recognition tasks)
• EEG preprocessing methods and artifact removal techniques
• Feature extraction approaches across time, frequency, and wavelet domains
• Classification algorithms ranging from classical machine learning to deep neural networks
• Reported accuracy rates and performance metrics

Results & Key Findings

CIT with ERP P300 was the most frequently employed protocol, reflecting the research community's convergence on recognition-based rather than arousal-based deception detection. The review identified clear methodological trends in how researchers are processing and analyzing EEG signals for lie detection.

For signal preprocessing and feature extraction, the findings revealed strong preferences for specific techniques. The most used preprocessing method was Bandpass Filtering (BPF), and the Discrete Wavelet Transform (DWT) emerged as the preferred feature extraction technique due to its suitability for non-stationary EEG signals. DWT's ability to capture both time and frequency information makes it particularly well-suited for analyzing transient events like the P300.

Among classification algorithms, Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), and Convolutional Neural Networks (CNN) were frequently utilized. Notably, an accuracy score of 99.88% was achieved using the LSTM and DWT approaches in one study reviewed, demonstrating the exceptional performance potential of hybrid deep learning architectures.

Key methodological findings included:

• Bandpass filtering as the dominant preprocessing approach for noise reduction
• DWT-based feature extraction outperforming traditional time-domain or frequency-domain methods
• Hybrid and deep learning models (CNN-LSTM combinations) achieving superior classification performance
• Face recognition paradigms emerging as particularly effective stimulus types for eliciting robust P300 responses

These findings demonstrate the effectiveness of hybrid and deep learning-based models in enhancing classification performance. The convergence on specific methodologies across independent research teams suggests these approaches have achieved a level of maturity and reliability.

Discussion & Significance

EEG-based lie detection, particularly using the ERP P300 response to face recognition tasks, shows promising accuracy and robustness compared to traditional polygraph methods. This represents a fundamental advancement: unlike peripheral physiological measures that index general arousal or stress, the P300 reflects cognitive recognition processes that are involuntary and difficult to suppress.

Combining advanced signal processing methods with machine learning and deep learning classifiers significantly improves performance. The review's identification of DWT as the preferred feature extraction method is particularly significant—it suggests the field has moved beyond simple amplitude measurements to sophisticated multi-resolution analysis that captures the complex temporal dynamics of deception-related brain activity.

The dominance of the CIT protocol reflects a theoretical shift in deception detection: from attempting to detect the stress of lying to detecting the presence of concealed knowledge. This cognitive approach is inherently more specific and less vulnerable to false positives from general anxiety or stress. The effectiveness of face recognition paradigms is especially noteworthy for real-world applications, as faces are highly salient stimuli that reliably elicit strong P300 responses and are directly relevant to many investigative scenarios.

The review's findings position EEG-based deception detection as a viable alternative or complement to traditional polygraphy. With accuracy rates approaching or exceeding 99% using optimized hybrid models, these methods are beginning to meet the stringent reliability standards required for forensic and security applications. The ability to directly measure brain activity associated with recognition rather than relying on indirect peripheral measures represents a qualitative leap in lie detection technology.

Limitations & Considerations

Despite the impressive accuracy figures reported, the review inherently reflects the limitations of the primary literature. Most reviewed studies were conducted in controlled laboratory settings with relatively small sample sizes and mock-crime scenarios. The ecological validity of these findings for real-world interrogations, field investigations, or high-stakes security screening remains to be fully established.

The review focused exclusively on English-language publications from four major databases, potentially missing relevant research published in other languages or venues. Additionally, publication bias likely favors studies reporting positive results and high accuracy rates, potentially inflating the apparent effectiveness of these methods.

Important methodological concerns include:

• Limited evidence on performance with sophisticated countermeasures or trained deceivers
• Variability in how different studies define and operationalize "deception"
• Insufficient data on cross-cultural validity and individual difference factors
• Computational complexity of hybrid deep learning models limiting real-time deployment

The field also lacks standardization in electrode placement, stimulus presentation protocols, and data analysis pipelines, making direct comparisons across studies challenging. While the review identifies best practices, it cannot resolve these fundamental standardization issues that plague the broader neurophysiological deception detection field.

Practical Applications

This systematic review provides evidence-based guidance for practitioners and researchers seeking to implement EEG-based lie detection systems. The clear recommendation to use bandpass filtering for preprocessing, DWT for feature extraction, and hybrid CNN-LSTM architectures for classification offers a validated methodological pathway for system development.

For forensic and security professionals, the review's findings suggest that EEG-based CIT using face recognition paradigms could complement traditional polygraphy in criminal investigations. The technology may be particularly valuable for identifying suspects who possess knowledge of crime details or for verifying alibis through recognition testing. However, implementation should proceed cautiously given the laboratory-to-field generalization gap.

For researchers, the review highlights critical future directions: developing real-time processing capabilities, testing cross-individual generalization, evaluating countermeasure resistance, and reducing system complexity to facilitate broader adoption. The field is poised for translation from research to practice, but requires additional validation in operational contexts before widespread deployment in high-stakes decision-making.