Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here

Hong Hanh Nguyen-Le

Project Title

Enhancing the Robustness and Efficiency in Biometric Authentication Systems using Representation Learning and Fuzzy Signature Scheme

Project Description

Recently, biometric-based authentication systems (BASs) are a promising alternative to password-based authentication systems since they escape people from memorizing long passwords or carrying physical devices. Existing privacy-preserving approaches in biometric authentication systems (BASs) are the aggregation between DL-based feature extractor and biometric template protection (BTP) scheme. However, these methods have several drawbacks: 1) To derive a cryptographic key (CK), most approaches are building DL models to map each user’s feature to a predetermined CK, leading to the inflexibility of those models in cases of any changes in key length being occurred. 2) Most current BTP schemes require the usage of helper data to reconstruct protected template, which can reveal users’ confidential information. 3) DL models are vulnerable to adversarial attacks, negatively affecting the overall security of BASs. Therefore, this research aims at developing an end-to-end (E2E) Deep Feature Extractor (DFE) which directly encodes the biometrics features into unique hashing codes, and improving the security of BTP scheme using fuzzy signature (FS) scheme.