SpamShield

unsafe(HTTP) url detection
shortend url detection
phishing link detection
spam mail detection
valid mail detection
scam detector using screenshot

Inspiration

As everyone millions got scammed from fake emails and phishing links, we wanted to create a universal shield against spam threats.

What it does

SpamShield AI is a comprehensive multi-channel spam detection system that protects users across three critical threat vectors:

Email Spam Detection using Machine Learning (Naive Bayes)
Phishing Link Detection using rule-based analysis
Malicious Screenshot Detection using OCR + ML

How we built it

Built with Python, Streamlit, Scikit-learn, and Tesseract OCR. We trained a Naive Bayes classifier on spam datasets, implemented 4 sophisticated link detection rules (HTTP, domain structure, phishing keywords, shortened URLs), and created an OCR pipeline for image analysis.

Challenges we ran into

Balancing accuracy vs false positives in ML model
Handling diverse phishing patterns and evolving threats
Integrating multiple detection modules into one seamless UI

Accomplishments that we're proud of

Created first unified 3-channel spam detection system
Achieved high accuracy across all three detection modules
Built production-ready Streamlit interface
Deployed working prototype with real-time analysis
Scalable architecture ready for API integration

What we learned

Machine Learning models need diverse training data to handle real-world spam variations

Rule-based detection complements ML perfectly for structured threats like URLs
OCR technology is powerful but requires careful text preprocessing for accuracy
User experience is critical - a complex system needs a simple, intuitive interface
Multi-channel approach catches threats that single-method systems miss
Balancing false positives vs false negatives is the key challenge in security