Erinyes: A Continuous Authentication Protocol
Abstract
The need for user authentication in the digital domain is paramount as the number of digital interactions that involve sensitive data continues to increase. Advances in the fields of machine learning (ML) and biometric encryption have enabled the development of technologies that can provide fully remote continuous user authentication services. This thesis introduces the Erinyes protocol. The protocol leverages state of the art ML models, biometric encryption of asymmetric cryptographic keys, and a trusted third-party client-server architecture to continuously authenticate users through their behavioral biometrics. The goals in developing the protocol were to identify if biometric encryption using keystroke timing and mouse cursor movement sequences were feasible and to measure the performance of a continuous authentication system that utilizes biometric encryption. Our research found that with a combined keystroke and mouse cursor movement dataset, the biometric encryption system can perform with a 0.93% False Acceptance Rate(FAR), 0.00% False Reject Rate (FRR), and 99.07% accuracy. Using a similar dataset, the overall integrated system averaged 0% FAR, 2% FRR and 98% accuracy across multiple users. These metrics demonstrate that the Erinyes protocol can achieve continuous user authentication with minimal user intrusion.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2022
- Accession Number
- AD1201035
Entities
People
- Fabian A. Lopez
- Gage A. Hathaway
Organizations
- Naval Postgraduate School