Modification of the 5D Lorenz chaotic map with fuzzy numbers for video encryption in cloud computing

Abstract

As surveillance cameras have proliferated in usage, their widespread deployment has raised privacy concerns. We introduce an inventive strategy to safeguard privacy in surveillance videos to address these concerns. This article designs a secure system for detecting and encrypting regions of interest (ROIs) that depict multiple individuals within video footage. The suggested system is composed of three phases, with the initial phase incorporating an object detection model to efficiently detect individuals in video frames with the You Only Look Once version 7 architecture. The second stage encrypts ROIs with our unique algorithm, which represents a novel technique derived from combining triple DNA with the modification of the 5D Lorenz chaotic map using fuzzy triangular numbers, which are utilized in key generation. The reverse of this process is a decryption that obtains the original video. The third stage combines all encrypted ROIs from the reconstructed video frames to be securely stored as encrypted video in the cloud. Evaluation results show that the utmost value of the unified averaged changed intensity and the number of changing pixel rate stand at 33.8000 and 99.8934%, respectively, with encryption and decryption speeds up to 7.06 and 6.72 s, respectively.