A Dynamic Obfuscation Framework for Security and Utility
Andrew Wintenberg, Matthew Blischke, Stéphane Lafortune, Necmiye Özay · 2022
Obfuscation can be used by dynamic systems to ensure private and secure communication over networks vulnerable to eavesdroppers. Balancing the utility of sending information to intended recipients and privacy by hiding information from unintended recipients presents an interesting challenge. We propose a new framework for obfuscation that includes an inference interface to allow intended recipients to interpret obfuscated information. We model the security of the obfuscation with opacity, a formal notion of plausible deniability. Using techniques from distributed reactive synthesis, we show how to automatically design a privacy-enforcing obfuscator along with the inference interface that is given to intended recipients to use as a “key”. We demonstrate this approach by enforcing privacy while maintaining utility in a contact tracing model.