This powerful AI fraud detection engine identifies and flags various risk content such as scams, phishing, fake job postings, and gambling advertisements. It offers robust bilingual support, automatically adapting to English and Chinese texts using specialized models. With its hybrid detection approach, combining LightGBM expert models and Chinese keyword rules, it delivers real-time responses with risk scores, determinations, and severity levels. Integrated with the MCP protocol and a REST API, it can be directly called by AI clients like GitHub Copilot, ensuring content safety and offering interpretability through fraud probability scores.
Características Principales
01Real-time responses with risk scoring and determination
02Hybrid detection combining LightGBM and keyword rules
03Outputs fraud probability for interpretability
04Bilingual support for English and Chinese text
05MCP protocol integration for AI Agent calls
060 GitHub stars
