FACULTY OF ENGINEERING COSIC PrivacyPreserving Multibiometric Authentication in

FACULTY OF ENGINEERING COSIC Privacy-Preserving Multibiometric Authentication in Cloud Authors: Christina-Angeliki Toli, Abdelrahaman Aly and Bart Preneel KU Leuven, ESAT/COSIC and imec Euro S&P 2017 2 nd IEEE European Symposium on Security and Privacy Paris, France 27 April 2017

Motivation § Cloud-based biometric applications § Privacy-by-design approaches § Link the pattern recognition area with applied crypto § Biometrics integration is still an “open problem” § Addressing the topic from a multimodal perspective, taking into account impacts on performance § Secure Multiparty Computation (MPC) 2

Layout Figure: Multimodal Biometric Authentication in Cloud Environment 3

Contribution § Re-use the prior stored unimodal templates, avoiding the auxiliary temporary or permanent storage § A user-specific performance-dependent weighted score level fusion strategy § MPC protocols for cloud in a privacy-preserving decentralized manner § Threats from parties with competing interest, collusion models, security and privacy perspectives § Entity learns only the {0, 1} value of the final result 4

Resutls § Authentication/identification by setting up a FAR or FRR-dependent fusion § The operation is dynamic § Stored templates, matching, normalized and fusion scores are inaccessible § Privacy of user's information from the output party § Limitation: Identification is time-consuming processes § Future work: Fusion accuracy, computational and communication complexity and e�ciency 5

Thank you for your attention ! Contact : http: //www. esat. kuleuven. be/cosic/ Christina-Angeliki Toli ctoli@esat. kuleuven. be