News Credibility Analyzer

Fake News Detector helps you browse smarter by instantly flagging satirical, conspiratorial, and misleading news—without sacrificing your privacy. Built for real-time use, the extension combines curated domain knowledge, pattern matching, and a lightweight machine learning ensemble to surface questionable content in under 50ms—all locally in your browser. No data is ever sent to servers: everything runs client-side, preserving your browsing privacy. Key Features: Real-Time Analysis: Articles are evaluated as you browse (≈29ms per page) so you get instant signals. Multi-Type Detection: Differentiates satire, conspiracy theories, medical misinformation (e.g., “miracle cure” claims), and political misinformation. Domain Recognition: Immediately flags known parody/conspiracy domains (e.g., The Onion, Babylon Bee, InfoWars) with high confidence. Pattern Matching: Detects linguistic cues through curated fake/satirical/real news patterns to supplement the ML signal. Ensemble Machine Learning: Combines Logistic Regression and Naive Bayes with TF-IDF n-gram features, optimized down to 295 top signals for fast in-browser inference. Privacy-First: All inference happens locally—no network requests, no tracking, no data collection. Lightweight: Browser model is small (~500KB) with aggressive optimization for speed and memory. Confidence Scoring & Thresholding: Only surface warnings when multiple signals align (configurable confidence cutoff). Rich UI Feedback: Color-coded indicators and banners (satire, misinformation, reliable) with one-click dismiss and auto-hide. How It Works: Domain Check: Fast lookup against known problematic sites for immediate classification. Pattern Analysis: Regular expressions scan article text for known fake/satirical phrasing. Statistical Inference: A trimmed-down ensemble ML model processes the cleaned text to provide a probabilistic score. Fusion: Signals are weighted (domain, pattern, and ML) to compute an overall confidence and trigger a user-visible alert if warranted. Why Use It: Stay better informed by spotting misleading or parody content before you share it. Avoid echo chambers and accidental spread of misinformation. Keep your data private—no external APIs, no telemetry. Lightweight and fast; won't slow down your browsing. Limitations & Transparency: Trained primarily on political/news data from the 2016 election era—some emerging topics may require retraining. Conservative design favors recall; occasional legitimate articles may be flagged (false positives), but multi-layer validation reduces noise. Getting Started: Install and enable the extension. Browse news sites normally. Watch for the popup indicator or banner if content is flagged—hover or click for details. Open Source & Auditable: The entire system is transparent and can be reviewed or contributed to on GitHub. Support & Contribution: Fork the repo, open issues, or submit improvements via pull requests.