AI-Based Auction Trust and Fraud Detection Platform: A Predictive System for Manipulated Ratings and Suspicious Transactions in Nigerian Online Marketplaces

Abstract

Online auction and marketplace platforms such as eBay, Amazon Marketplace, Jumia, and Konga sit at the centre of digital commerce, yet remain exposed to trust-eroding fraud, including fake listings, cloned product images, manipulated ratings, shill bidding, and fraudulent sellers. Within Nigeria, these trust deficits form a substantial barrier to marketplace adoption, a problem worsened by the fact that most commercial fraud-detection tools are trained on Western data and cannot reliably parse Nigerian Pidgin, Yoruba, Igbo, or Hausa code-switching in local listings and reviews. This paper presents the design of TrustBid AI, a predictive artificial-intelligence system built to counter four fraud modalities on Nigerian auction platforms: manipulated ratings, fake reviews, cloned product images, and suspicious bidding activity. The architecture rests on four cooperating components: behavioural modelling, content analysis, network interaction signals, and human feedback loops, fused into a hybrid model that reads text, images, and behavioural sequences together rather than in isolation. Behavioural modelling flags shill bidding and bid shielding through sequence-based anomaly detection; content-analysis models score ratings, reviews, and product photographs for manipulation; network analysis maps buyer–seller relationships to expose coordinated fraud rings; and a moderator-in-the-loop layer lets verified reviewers correct and refine model output. Outputs from the four layers are combined into weighted Seller Trust, Item Trust, and Image Authenticity scores rendered on a buyer-facing mobile interface and a moderator web console, both of which are illustrated with interface mock-ups in this paper. A functional prototype is proposed for evaluation on locally sourced Nigerian marketplace data, governed by the Nigeria Data Protection Regulation and its successor, the Nigeria Data Protection Act 2023, with a companion framework for fraud-intelligence sharing among Nigerian platform operators. The expected outcome is a deployable, language-aware trust layer that improves fraud-detection accuracy, reduces buyer exposure to deceptive content, and strengthens confidence in Nigeria's growing digital economy.

Keywords

References