User experience and robustness in social virtual reality applications

Abstract

Cloud-based applications that rely on emerging technologies such as social virtual reality are increasingly being deployed at high-scale in e.g., remote-learning, public safety, and healthcare. These applications increasingly need mechanisms to maintain robustness and immersive user experience as a joint consideration to minimize disruption in service availability due to cyber attacks/faults. Specifically, effective modeling and real-time adaptation approaches need to be investigated to ensure that the application functionality is resilient and does not induce undesired cybersickness levels. In this thesis, we investigate a novel 'DevSecOps' paradigm to jointly tune both the robustness and immersive performance factors in social virtual reality application design/operations. We characterize robustness factors considering Security, Privacy and Safety (SPS), and immersive performance factors considering Quality of Application, Quality of Service, and Quality of Experience (3Q). We achieve �harmonized security and performance by design" via modeling the SPS and 3Q factors in cloud-hosted applications using attack-fault trees (AFT) and an accurate quantitative analysis via formal verification techniques i.e., statistical model checking (SMC). We develop a real-time adaptive control capability to manage SPS/3Q issues affecting a critical anomaly event that induces undesired cybersickness. This control capability features a novel dynamic rule-based approach for closed-loop decision making augmented by a knowledge base for the SPS/3Q issues of individual and/or combination events. Correspondingly, we collect threat intelligence on application and network based cyber-attacks that disrupt immersiveness, and develop a multi-label K-NN classifier as well as statistical analysis techniques for critical anomaly event detection. We validate the effectiveness of our solution approach in a real-time cloud testbed featuring vSocial, a social virtual reality based learning environment that supports delivery of Social Competence Intervention (SCI) curriculum for youth. Based on our experiment findings, we show that our solution approach enables: (i) identification of the most vulnerable components that impact user immersive experience to formally conduct risk assessment, (ii) dynamic decision making for controlling SPS/3Q issues inducing undesirable cybersickness levels via quantitative metrics of user feedback and effective anomaly detection, and (iii) rule-based policies following the NIST SP 800-160 principles and cloud-hosting recommendations for a more secure, privacy-preserving, and robust cloud-based application configuration with satisfactory immersive user experience.