What Inspired Us
The idea for PhantomClick came from real-world scam incidents where even tech-savvy users lost money by clicking links that looked completely legitimate. Most scam warnings today are vague and reactive, offering no clarity or proof. We wanted to build a system that doesn’t just warn but actually shows users what a scam does safely and transparently.
What We Learned
Through this project, we gained hands-on experience with scam psychology, phishing techniques, and real-world fraud patterns. Technically, we learned how to combine AI-driven text analysis, URL risk scoring, sandboxed browser execution, and automated forensic reporting into one cohesive pipeline. We also learned how critical explainability is in security tools users trust systems that show evidence, not just alerts.
How We Built It
PhantomClick was built as a multi-stage pipeline. We extract text and links from SMS screenshots using OCR, analyze intent and urgency using AI models, and score URLs using heuristic and machine-learning techniques. Suspicious links are then opened inside a secure, isolated sandbox where their behavior is monitored. Finally, we generate visual proof, scam replays, and structured forensic reports ready for cybercrime reporting.
Challenges We Faced
Our biggest challenges were safely interacting with malicious links without exposing users, reducing false positives, and translating complex technical behavior into simple, human-readable insights. Balancing security, accuracy, and usability required multiple design iterations.
Accomplishments that we’re proud of
We’re proud of building a system that goes beyond detection and delivers real, explainable proof. Creating automated, court-ready forensic reports and scam replays that visually explain an attack was a key milestone for us.
What we learned
We learned how modern scams combine technical exploits with psychological manipulation. The project also helped us understand the importance of explainability, user trust, and secure system design in cybersecurity tools.
What’s next for PhantomClick
Next, we plan to improve model accuracy, expand scam detection coverage, and integrate PhantomClick with banking apps, telecom SMS gateways, and cybercrime investigation workflows to stop scams earlier and at scale.
Built With
- axios
- express.js
- fastapi
- gemini
- node.js
- pil/pillow
- playwright
- pytesseract
- react
- tailwindcss
- tesseract
- vite
Log in or sign up for Devpost to join the conversation.