LLM-Augmented Adaptive Security Control for Virtual Reality Streaming

Abstract

Live Virtual Reality (VR), streaming has stringent motion-to-photon latency demands and it is highly susceptible to transport-layer and network-level cyber attacks. Despite the fact that most protocols such as protocols such as TLS 1.3 or DTLS can provide high level of cryptographic assurances, their fixed implementation has a performance cost that can make immersive interaction in dynamic adversarial environments unstable.The current paper provides a protocol that is a latency-constrained LLM-enhanced adaptive transport-layer security protocol named LARS-VR, which is designed to support real-time VR streaming. The threat intensity, cryptographic overhead and end-to-end latency are modeled as dynamically coupled state variables on a constrained discrete time control architecture. Security adaptation is formulated as a finite-time constrained search problem over a finite admissible action set, as immersive latency feasibility is a system-wide property that is strictly enforced. The proposed architecture will guarantee limited threat evolution and stable immersive performance without having encrypted data and modifying cryptographic primitives. Conditional simulation and analytical formulation under Wi-Fi 6 conditions display as much as 18 per cent reduction in mean latency, 15-20 per cent increase in effective throughput and a substantial reduction in latency variance compared to fixed secure setups. These results constitute a mathematically constructed paradigm of cognitive and latency-conscious transport-layer security in immersive VR and other systems that are sensitive to delay.