AI Scam Detector Chat Assistant

[Mrinalray]

CyberShield currently focuses on URL scanning, but many online scams begin through messages, emails, social media posts, job offers, OTP requests, and payment requests before users even click a link.

This feature introduces an AI-powered Scam Detector Chat Assistant where users can paste suspicious messages, emails, screenshots, or conversations and receive an instant scam analysis with explanations.

The assistant should identify common scam patterns such as phishing attempts, fake job offers, lottery scams, tech support scams, investment fraud, impersonation attacks, and social engineering tactics.

[aayushprsingh]

Hi, I am a GSSoC 2026 contributor and would like to work on this. this project is my primary stack. Let me know if it's available!

Here's how I'm thinking of approaching this:
I'll implement a dedicated chat endpoint leveraging the Gemini API with a system prompt specifically weighted for detecting social engineering markers and impersonation patterns. I also plan to integrate an OCR layer to handle screenshots, ensuring we can parse text from visual evidence before running it through the scam-likelihood classification model.

[Antra1705]

I'll build a ScamDetectorChat React component with a message input area (supporting text paste), wire it to a new /api/scam-detect backend endpoint that sends the user's message to an LLM with a structured system prompt covering phishing, fake job offers, OTP fraud, investment scams, and impersonation patterns — returning a structured response with threat classification, confidence level, and explanation. The chat UI will maintain conversation history for follow-up questions.
My ZerodhaClone handles real-time message-driven UI in React, and ApiForge handles structured API request/response flows — both patterns apply directly to building this chat interface and backend endpoint.
Contributing under GSSoC '26. Could you please assign this to @Antra1705?

[Mrinalray]

@Antra1705 assigned