Fraud App Detection Using Sentiment Analysis and Machine Learning

Abstract

The rapid expansion of mobile applications has transformed the way users access digital services, including banking, healthcare, education, entertainment, and electronic commerce. Along with this rapid growth, fraudulent and malicious mobile applications have become increasingly prevalent, posing serious threats to user privacy, financial security, and personal information. Fraudulent applications often imitate legitimate applications, request unnecessary permissions, steal confidential data, or engage in deceptive activities that negatively affect users. Traditional app verification methods primarily rely on manual inspection and static security analysis, which are often insufficient for identifying newly emerging fraudulent applications. This research proposes an intelligent fraud app detection system based on sentiment analysis and machine learning techniques. The proposed system collects user reviews from application stores and applies Natural Language Processing (NLP) methods to preprocess textual data through tokenization, stop-word removal, stemming, and vectorization. Sentiment analysis is performed to identify positive, negative, and neutral opinions expressed by users, while machine learning algorithms classify applications as genuine or fraudulent based on extracted textual features. Experimental evaluation demonstrates that sentiment-based classification significantly improves fraud detection accuracy by identifying suspicious behavioral patterns hidden within user reviews. The proposed approach provides a scalable, cost-effective, and automated solution for protecting users against fraudulent applications and enhancing mobile application security.

Keywords