Signature Forgery Detection

This project began as an attempt to improve digital signature authentication by creating a system that can identify both the visual structure of a signature and the way it is written. I was inspired by the challenge of reducing fraud while maintaining fairness for genuine users. I built the system by capturing behavioral data such as timing, curvature, pressure, and stroke order, and combining those features with static shape analysis to create a more reliable similarity score. One of the biggest challenges was preventing false positives on simple signatures, which are easy to forge, and I solved this by tightening geometric comparisons and introducing adaptive thresholds that become more strict when the signature has low complexity. I also worked to avoid false negatives on more complex signatures by adding writer-specific tolerance adjustments. Throughout the project, I learned about biometric pattern analysis, feature engineering, and the balance required to build a verification system that can detect forgery while still accounting for natural human variation.

Built With

• css
• javascript
• typescript