Background

With the exponential rise in digital content creation and social media sharing, the spread of fake news, manipulated images, and deepfake videos has become a major threat to public trust, national security, and social harmony. Coordinated misinformation campaigns using AI-generated content can incite violence, manipulate elections, ruin reputations, and create panic in society.

Conventional content moderation techniques struggle to keep up with the volume, speed, and sophistication of modern disinformation. There is an urgent need for AI-powered tools that can identify and verify suspicious content across formats in real-time or near-real-time.

Problem Statement

No unified, user-friendly platform currently exists that can automatically and accurately detect and analyse textual fake news, AI-generated deepfake videos, and digitally manipulated images, especially in Indian languages or localized content. Media literacy is also lacking, leaving citizens vulnerable to deception.

Key Objectives

• Equip users with a reliable verification tool
• Prevent the viral spread of fake news and deepfakes
• Promote responsible digital consumption and reporting
• Support law enforcement and fact-checkers with scalable technology

Functional Requirements

Develop an AI-Based Digital Media Verification Tool (web or mobile-based) that can:

 Fake News Detection (Text-Based)

o Analyse news articles, social media posts, and headlines to:

o Detect clickbait, misleading claims, or known hoaxes

o Cross-verify with trusted fact-checking databases (e.g., PIB, AltNews, Boom)

o Use NLP techniques for stance detection and semantic similarity

 Deepfake Video Detection

o Accept user-uploaded videos and perform:

o Face tracking and head movement analysis

o Audio-lip sync detection

o Temporal and spatial artifacts identification

o Use pretrained deepfake detection models (e.g., MesoNet, Deepware)

 Image Manipulation Detection

o Accept user-uploaded images and scan for:

o Photoshop and AI-generated artifacts

o Error Level Analysis (ELA), metadata inconsistencies

o GAN fingerprints (for AI-generated content)

Additional Requirements

• Provide a trust score or authenticity rating
• Give users explanations or visual cues for flagged content
• Support multiple Indian languages (Hindi, Gujarati, Tamil, etc.)
• Educate users through media literacy tips and awareness modules

Evaluation Criteria

• Accuracy and reliability of detection (low false positives/negatives)
• Technical innovation and use of AI/ML
• Speed and scalability of the solution
• User interface and experience
• Language support and regional adaptability

Bonus Points

• Real-time browser or social media plugin
• Integration with news verification APIs or media databases
• Crowd-sourced flagging and reporting mechanisms
• Offline capability for remote regions
• Dashboard for fact-checking teams or law enforcement