Neural cellular automata approach for reducing security and privacy hazards of augmented reality environment

The applications of the Augmented Reality (AR) environment are getting traction in the field of healthcare, smart city navigation, and industrial training and collaboration working environments. Nevertheless, this constant streaming of multimedia and the exchange of sensor information may subject the security and privacy threat of illegitimate access, data capture, and inference. Unlike prior AR frameworks, the proposed study incorporates a novel Neural Cellular Automata (NCA) layer that actively learns AR behavior and dynamically enforces spatial-temporal anomaly detection, adaptive access control and encryption. The proposed NCA layer leverages the localized state evaluation to define the spatial dependence and distributed behavioral pattern in the AR stream. The NCA layer defines rules for each stream and monitors the AR environment activities elegantly, anomaly detection with adaptive neural models, in its dynamic access decision. The layer further incorporated with an optimized key generation algorithm on a symmetric encryption technique to ensure secure transmission with minimal computational overhead. The experimental results of the proposed study show the intrusion detection rate is 96.8%, access control decision accuracy is 94.3%, and the overhead of encryption was below 6%. Analysis of a deployment in the post-deployment stage indicated that this multi-layered mechanism of privacy lessened measurable AR security and privacy risk by more than 85% in simulated attack conditions.