Safesurf

SafeSurf: AI-Based Web Security Engine

Comment

🚀 SafeSurf – AI-Powered Browser Safety System

Inspiration

With the increasing number of phishing attacks, malicious websites, and emerging threats like prompt injection, users are more vulnerable than ever while browsing. Most existing solutions are either reactive or lack real-time intelligence.

SafeSurf was inspired by the need to create a proactive, intelligent, and real-time browser security system that can protect users from both traditional and modern web-based threats.

What it does

SafeSurf is a browser extension that enhances user safety by detecting and preventing multiple types of threats:

  • 🔗 Detects phishing URLs using PhishTank
  • ⚠️ Identifies suspicious URL structures (e.g., obfuscation)
  • 📄 Flags low-content or deceptive web pages
  • 🧠 Detects prompt injection attempts
  • 🔐 Identifies potential API key leaks or sensitive data exposure
  • 🚨 Provides real-time alerts with options like Go Back or Continue

🛠️ How we built it

We built SafeSurf as a Chrome Extension using:

  • JavaScript for core logic and browser interaction
  • PhishTank API for real-time phishing URL detection
  • Custom heuristics for analyzing URL patterns and page content
  • A dynamic alert UI overlay to notify users instantly
  • Pattern-based detection for prompt injections and API leaks -Integrated gemini-2.5-slash to analyze page

The detection pipeline is modular, allowing future integration of AI/ML models and LLMs.

⚡ Challenges we ran into

  • Reducing false positives in heuristic-based detection
  • Balancing security and usability (giving users control with warnings) -Working with gemini API it only allows 20 calls per session which is insuitable for chrome entensions i'm thinking to replace gemini api with Groq's API which is free and also provide around 14000 calls ---

Accomplishments so far

  • Built a working prototype capable of detecting phishing sites live
  • Implemented multi-layered threat detection (URL + content + behavior)
  • Added detection for prompt injection and API leaks
  • Designed a clean and intuitive threat alert system
  • Created a strong foundation for future AI-powered enhancements

📚 What we learned

  • How browser extensions interact with web pages and APIs
  • Practical challenges in building real-world cybersecurity systems
  • Importance of layered security (Defense in Depth):

$$ \text{Security} = \text{URL Analysis} + \text{Content Analysis} + \text{Behavior Detection} $$

  • How to integrate external threat intelligence sources like PhishTank
  • Designing user-friendly security experiences

🔮 What's next for SafeSurf

  • Integrate LLMs (Ollama) for intelligent threat reasoning
  • Develop an offline ML model for phishing detection
  • Add cookie tracking and privacy management
  • Build a dashboard for threat analytics
  • Expand detection to malware, trackers, and unsafe downloads
  • Deploy SafeSurf as a full-scale browser security product

Built With

Share this project:

Updates