SpamShield

Inspiration

I've personally received several phishing emails, and I've seen people around me fall for phishing attempts. Seeing how fast phishing involved in 2025, especially with attackers using AI - inspired me to build smarter system called SpamShield

What it does

It analyzes emails using an LLM, checks sender authenticity, detects dangerous attachments, and automatically move high-risk messages to a QUARANTINE folder.

How i built it

The MVP is a Python script using imap, basic attachment(extensions) inspection, and an external LLM classifier. The model evaluates message intent, patterns and writing style, while traditional checks usualy validate only SPF/DKIM/DMARC and domain/ip anomalies.

Challenges i ran into

High false-positives rate during early testing Handling tricky attachments formats and double extensions

Accomplishments that i'm proud of

Creating a functional LLM-powered filter Detecting sophisticated phishing messages that bypass traditional filters Building a clear quarantine system with human-readable explanations Designing an approach adaptable to new scam trends

What i learned

I learned how modern email authentication works, how to analyze suspicious attachments and how to use LLMs for classification

What's next for SpamShield

Add local pre-trained LLM to reduce dependency on external apis Implement the solution into mail client Automate reporting and security team responses Expand sender pattern analysis for more accurate threat detection

Built With

• imap
• mistralapi
• python