Fuzzy-logic threat classification for multi-level selective encryption over real-time video streams

Abstract

This paper proposes a Fuzzy-logic Threat Classification (FTC) model as the basis of a method to auto-detect three different confidentiality levels for videos streamed from heterogeneous, mobile devices via web edge servers, possibly part of a Content Distribution Network (CDN). The FTC consists of three parallel Fuzzy Inference Systems (FIS) corresponding to device, network, and type of video application, for the real-time, intelligent selection of an appropriate confidentiality level for a specific end-user. After selection of the level, an encryption module implements the corresponding form of encryption. In tests to demonstrate the concept, there were three increasing confidentiality levels, namely (1) low-level with no encryption, (2) Medium level with an in-house cipher [variant of eXclusive OR (XOR)], named P-XOR (XOR with additional rounds of permutation) applied to Selective Encryption (SE) and (3) high level with the Advanced Encryption Standard again for SE of compressed video syntax components. Results were obtained by considering realistic specifications of multiple digital devices, networks, and differing real-time streaming applications. Visual analysis of encrypted test video clips established that the FTC outputs an appropriate privacy level by reason of the implemented FISs. Absolute encryption times across the privacy levels were distinguished by their real-time response level, which is proportionate to the required degree of confidentiality.