SafeBiz: A Secure Framework for Safe Transactions in Social Media Marketplaces

Abstract

Social media sites are now real and companies can utilize them as a potential method of advertising their products and services. However, unscrupulous businesses are taking advantage of these channels to swindle their customers through false ads, viral links, and fake reviews. The conventional scamming detection systems are primarily focused on the text classification or profile analysis, and it usually fails in the case of fraud companies that mimic the legitimate online activity. The paper proposes a multi-signal machine learning system in identifying business scams in social media by integrating the profile credibility analysis, advertisement content analysis, review authenticity detection, website safety evaluation, and malicious URL detection. Each of the signals captures various dimensions of online credibility. Combining these heterogeneous signals through a weighted system of credibility fusion model gives the final Business Credibility Score (BCS) which indicates the likelihood of a business as legit or fraudulent and the experiments done with different public datasets such as Instagram fake account datasets, malicious URL datasets, and review datasets demonstrate that the proposed system has good detection results with an average F1-score of over 0.9593. The findings confirm that the application of composite signals enhances the identification of advanced social media frauds.