CyberCheckAI

cyberckeckai

Inspiration

With the rapid growth of AI-powered services, cyber threats are becoming more sophisticated and harder to detect. Phishing links, malicious URLs, and social engineering attacks are still one of the most common ways users get compromised.

The idea for this project came from the desire to create an AI-assisted security layer that can quickly analyze URLs and help users understand potential risks before interacting with them. I wanted to explore how modern AI models can be combined with classical security heuristics to make threat detection more accessible.

What it does

The project analyzes a given URL and produces a risk score and verdict based on multiple factors, such as:

Suspicious URL structure and patterns
Embedded credentials or obfuscation techniques
Redirect chains and final destination analysis
Context-aware AI reasoning to explain why a URL might be risky

The goal is not only to flag dangerous links, but also to explain the reasoning, making it useful for both technical and non-technical users.

How I built it

The system is built as a modular pipeline:

URL preprocessing – normalization and structural analysis
Heuristic checks – detecting common phishing and abuse indicators
Redirect resolution – analyzing where the link actually leads
AI analysis – using Gemini to interpret signals and generate a final verdict

The backend logic is designed to be easily extensible, allowing new security checks or AI prompts to be added without rewriting the whole system.

Challenges I ran into

One of the main challenges was finding the right balance between false positives and real threats. Overly strict rules can label safe links as dangerous, while relaxed rules can miss actual attacks.

Another challenge was making the AI output concise, explainable, and consistent, so that users trust the result instead of seeing it as a “black box”.

What I learned

How to combine rule-based security logic with AI reasoning
How to structure an AI-assisted pipeline instead of relying on a single model call
The importance of explainability in security-related applications

This project helped me better understand how AI can augment cybersecurity tools rather than replace traditional approaches.