AuthenReview

Abstract

The increasing reliance on online reviews for decision-making in digital platforms has significantly impacted consumer behaviour and business credibility. However, this reliance has also made online ecosystems vulnerable to manipulation by fake reviewers—individuals or automated accounts that impersonate legitimate users to disseminate misleading reviews. These fake reviewers distort public opinion, damage business reputations, and erode trust in digital marketplaces. While many existing systems attempt to detect fake reviews, they often overlook the importance of identifying the actual entities responsible for producing them. To address this issue, a system named AuthenReview was developed, focusing on the identification of fake reviewers through a machine learning-based approach. The system integrates behavioural analytics, linguistic feature extraction, and sentiment evaluation to assess reviewer authenticity. A score-based matrix was implemented using parameters such as review frequency, historical activity patterns, sentiment consistency, and metadata verification, including profile picture analysis. The backend was developed using Python and Flask, while TF-IDF was used for vectorization and a Random Forest classifier was employed for prediction. The system was evaluated using datasets comprising labelled reviews from multiple online platforms. The model achieved a recall of 0.6 and an F1-score of 0.46, indicating a promising capability to detect fake reviewers, especially in capturing a significant portion of fraudulent users. Although there is room for improvement in precision and accuracy, the results demonstrate that the combination of reviewer behaviour and content features is an effective strategy for mitigating online review fraud. This system lays the foundation for further research into scalable, multilingual, and real-time fake reviewer detection.