Browsing by Author "Baaqeel, Hind Ahmad"

Now showing 1 - 1 of 1

Restricted
A Security Model for Adaptive Face Authentication Systems Using Deep Learning Techniques
(Imam Abdulrahman Bin Faisal University, 2020) Baaqeel, Hind Ahmad; Olatunji, Sunday
With the widespread of computing and mobile devices, authentication using biometrics has received greater attention. Although biometric systems usually provide reasonable solutions, the recognition performance gets negatively affected over time due to changing conditions and ageing of biometric data, which results in intra-class variability. Fortunately, adaptive biometric have delivered as a solution to the addressed issue in which the system gradually adapts to new changes in user biometrics. In this thesis, an adaptive model has been introduced using deep learning and feature injection techniques to identify users even with a long-time gap. The proposed system has shown enhanced performance when compared to state of the art models. Notably, any system that employs templates updating mechanism may be exploited by an attacker to compromise the stored templates, either to impersonate a specific client or to deny access to him. The most common attack to adaptive authentication called poisoning attack where a skilled attacker gradually introduce engineered templates to the adaptive system until the imposter’s template becomes the new user reference. This attack was identified by Biggio as the main threat to any adaptive system. In furtherance to the objective of this thesis; a novel security framework has been developed to protect adaptive biometric systems against such kind of attack. It also ensures the robustness of the adaptive process since it provided the capability of recovering successfully from the attack when the system is suspected of getting poisoned. The new security model is reliable enough to protect users’ integrity, availability and privacy.
3 0