Title :
Secure Modular Hashing
Author :
Abelino Jim?nez;Bhiksha Raj;Jose Portelo;Isabel Trancoso
Author_Institution :
Carnegie Mellon University, Pittsburgh, PA, USA
Abstract :
In many situations, such as in biometric applications, there is need to encrypt and “hide” data, while simultaneously permitting restricted computations on them. We present a method to securely determine the ℓ2 distance between two signals if they are close enough. This method relies on a locality sensitive hashing scheme based on a secure modular embedding, computed using quantized random projections, being a generalization of previous work in the area. Secure Modular Hashes (SMH) extracted from the signals preserve information about the distance between the signals, hiding other characteristic from the signals. Theoretical properties state that the described scheme provides a mechanism to threshold how much information to reveal, and is also information theoretically secure above this threshold. Finally, experimental results reveal that distances computed from SMH vectors can effectively replace the actual Euclidean distances with minimal degradation.
Keywords :
"Euclidean distance","Authentication"
Conference_Titel :
Information Forensics and Security (WIFS), 2015 IEEE International Workshop on
DOI :
10.1109/WIFS.2015.7368567