This project aims at advancing the state-of-the-art in cybersecurity by developing efficient methods for generating novel biometric signatures and performing active and continuous user authentication. Current authentication procedures typically occur once at the initial log-in stage and involve user proxies such as passwords and smart cards which suffer from several vulnerabilities. This research addresses these limitations by developing probabilistic, generative models to represent multimodal biometrics of every user in the system, so that any significant deviation from the user-specific model flags the presence of an imposter. This novel technique specifically targets mitigation of masquerading attacks which are particularly challenging to detect as they are mostly carried out by insiders familiar with the activity patterns of the authorized user. The approach enables identification of intruders before they can hijack a user session of an authorized individual who may have momentarily stepped away from his/her console.
The project also involves extensive usability tests for seamless integration of the authentication processes in real-life computer and network systems, thus ensuring that data about the behavior and performance of people using the network is fed-back and incorporated into future designs of the security protocols. The methods also readily extend to protecting both wired and wireless networks, and mobile devices.
|