VeriX-AI: A Fake News Detection Platform Combining Machine Learning and Real-Time News Verification

VeriX-AI is an open-source fake news detection system that combines machine learning classifiers with real-time news RSS cross-verification to provide credibility assessments for news content. This project aims to address the social problem of fake news spread in the era of information explosion, adopting a front-end and back-end separation architecture. Its core components include a front-end web application, back-end services, and an AI analysis engine, which have application value in multiple scenarios but also have certain limitations.

In the era of information explosion, fake news and misleading content have become serious social issues, affecting public opinion and decision-making. Traditional manual review struggles to handle massive amounts of content, creating an urgent need for automated intelligent detection tools. As an open-source project, VeriX-AI is not just a text classifier but a comprehensive detection platform combining machine learning and real-time news verification.

VeriX-AI adopts a front-end and back-end separation architecture:

• Front-end: A user-friendly web application deployed on Vercel, supporting input of news content and returning detection results.
• Back-end: A serverless service built with Node.js, responsible for processing requests, calling AI modules, and returning results.
• AI Analysis Engine: The core module, including machine learning classifiers (based on TF-IDF and Logistic Regression/Passive Aggressive algorithms) and a real-time news verification module (via Google/Bing News RSS cross-verification).

Machine Learning Classification Process

1. TF-IDF Vectorization: Convert text into numerical feature vectors to capture the importance of keywords.
2. Heuristic Rules: Identify suspicious vocabulary (e.g., hoax, plandemic), emotional language (all caps, exclamation marks), and conspiracy theory keywords.
3. Credibility Signals: Detect citations from authoritative institutions, standardized citation formats, and specific news elements.

Real-Time News Cross-Verification

• Process: Extract keywords → Query Google/Bing News RSS → Evaluate source credibility → Calculate verification score.
• Trusted Media List: International and Indian local authoritative media such as Reuters, BBC, AP, etc.
• Scoring Logic: Confirmed by 2+ trusted media → Real (high confidence); 1 → Partially verified (medium); None → Rely on ML but lower confidence; Search failed → ML confidence cap at 75%.

Result Types

• LIKELY REAL (Probably Real), LIKELY FAKE (Probably Fake), UNCERTAIN series (Insufficient Evidence).

Scoring Metrics

• Confidence (0-100, model certainty), Trust Score (comprehensive ML and cross-verification), Source Credibility (source trustworthiness).

Influencing Factors

Such as "confirmed by 2 trusted media", "presence of low-credibility language patterns", "inciting format", "conspiracy theory language", etc.

VeriX-AI is applicable to multiple scenarios:

• Individual Users: Verify social media content to avoid spreading fake news.
• Content Platforms: Integrate into review processes to mark suspicious content.
• News Organizations: Quickly verify leads to ensure accurate reporting.
• Educational Institutions: Serve as an auxiliary tool for media literacy education.

Current Limitations

1. Language Limitation: Mainly optimized for English; support for other languages is limited.
2. RSS Dependence: Real-time verification is restricted by third-party services.
3. Training Data Bias: The model performs poorly in identifying some content.
4. Inability to Detect Deepfakes: Only processes text.

Improvement Directions

Multilingual support, multimodal detection (images/videos), knowledge graph integration, user feedback mechanism, blockchain verification.

VeriX-AI improves the accuracy of fake news detection through the dual mechanism of machine learning and real-time news verification. In today's world where it's hard to distinguish between true and false information, this tool provides a technical solution, but cultivating public media literacy and critical thinking is fundamental. It is hoped that this open-source project can help combat fake news and purify the online environment.