Background

Telegram has become a popular platform for both legitimate communication and illegal operations due to its encrypted features, large group support, anonymity, and minimal moderation. It is increasingly used to promote activities like:

• Drug trafficking
• Child Sexual Abuse Material (CSAM)
• Sale of counterfeit documents
• Terrorist propaganda
• Illegal arms and wildlife trade
• Crypto and investment frauds
• Leaked data dumps (data breach forums)

Manual monitoring of Telegram is impractical due to the sheer volume of content and its closed-group architecture. An AI-driven automated system is urgently required to detect and flag suspicious groups for timely action.

Problem Statement

There is currently no scalable tool that can automatically detect, analyse, and categorize Telegram groups that promote illegal, harmful, or banned content in real-time.

Key Objectives

• Enable authorities and cybersecurity teams to track illicit digital communities
• Automate detection and reduce reliance on manual surveillance
• Enhance public safety by proactively addressing digital crime hubs

Functional Requirements

Build an AI-powered tool (web, mobile, or backend platform) that can:

Core Functionalities:

1.  Crawl and monitor public Telegram groups/channels

o Use Telegram APIs, bots, or scraping methods

o Index posts, media, and metadata

2.  Detect illegal activity indicators using AI/ML/NLP:

o Keyword patterns (drugs, fraud, CSAM, darknet links, etc.)

o Code language or emojis used for illegal trade

o Frequent mentions of suspicious URLs or crypto addresses

o Group descriptions, usernames, and tags

3.  Classify groups based on risk level:

o Safe / Suspicious / Confirmed Illicit

4.  Provide dashboard and alerts:

o Real-time flagging to law enforcement or monitoring units

o Keyword heatmaps, trend analysis

o Group behaviour analytics (e.g., sudden user spike, high-volume posting)

Optional Advanced Features

• Image/video classification for CSAM or counterfeit detection
• Language support (Hindi, Gujarati, Arabic, etc.)
• Fake user profile detection via profile picture/content similarity
• Integration with government or intelligence databases

Evaluation Criteria

• Accuracy in detecting illicit content
• Innovation in group behaviour analysis
• Technical scalability and performance
• Privacy-respecting data handling
• UI/UX for analysts or law enforcement

Bonus Points

• Support for regional languages and slang detection
• AI-based user sentiment clustering within groups
• Real-time notification system with actionable reports
• Integration with external crime reporting systems