Keystroke study models behavioral drift for continuous authentication
Original title: Deep Learning Based Behavioral Drift Modeling for Continuous Biometric Authentication
Fixed-text login reaches 3.77% EER; free-text monitoring reports 10.05% EER on Buffalo and GREYC data.
Researchers propose a two-level keystroke system: fixed-text login with temporal convolutional networks, then continuous free-text monitoring by a four-model ensemble. Detected behavioral drift returns the user to Level 1 re-authentication. Population-based training on the fixed-text path is meant to adapt to long-term changes in password typing. Reported results are 96%+ fixed-text accuracy and 83.53% true acceptance at 5% false acceptance for free text.
- Level 1 uses contrastive learning and population-based training.
- Free-text continuous authentication reports 83.53% TAR at 5% FAR.
- Evaluation uses the Buffalo free-text and GREYC fixed-text datasets.
- Suspicious drift sends the session back to Level 1.
From the original article
1 min read at arxiv.orgThis research proposes a two-level keystroke dynamics authentication system that provides defense-in-depth security through continuous behavioral drift monitoring. Level 1 employs fixed-text password authentication using Temporal Convolutional Networks (TCN) with contrastive learning and population-based training (PBT), achieving 3. 77% EER for high-security login verification.
By Suprava Saha Dibya · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.