Mr. Jay Bharat Mehta

Autonomous Patch Validation for Zero-day Exploits in Enterprise Clouds

Abstract:

Enterprise cloud infrastructures are constantly at risk from zero-day vulnerabilities, which frequently get past traditional security protections before efficient remedies can be put in place. Conventional patch validation techniques are usually slow, reactive, and lack the analytical depth needed to properly link patch interventions to results. This self-contained patch validation system integrates anomaly detection, predictive risk modeling, and causal inference into a single pipeline. The NSL-KDD dataset was used to train and assess machine learning models, enabling proactive prioritizing of high-risk patches, real-time anomaly detection, and statistical validation for change impact through causal analysis. The recommended method achieved a 93.76% detection accuracy and significantly reduced the detection time from 0.61s to 0.22s. Causal inference confirmed that the deployed patches were responsible for the reported behavioral changes with a 99.4% likelihood.

Ablation studies validated the contribution of each module, and unsupervised detection further enhanced the system's robustness. The technology provides a scalable, interpretable, and efficient method of addressing enterprise cloud zero-day vulnerabilities. By enabling quicker, more reliable, and understandable patch distribution through a combination of statistical analysis and causal validation, it enhances the overall security posture.

Profile:

Jay Bharat Mehta is a seasoned engineer and technology innovator with over nine years of experience in Enterprise Test Engineering, specializing in AI-enhanced security-critical automation and distributed cloud systems. He has contributed to world-class platforms at Snowflake Inc., Apple Inc., Guidewire Software, and Wells Fargo, driving large-scale quality and reliability initiatives that delivered multi-million-dollar savings. Projects he contributed to have been featured on Financial Post, Yahoo Finance, and Medium, including work on cyber risk modeling, eSIM systems, and scalable data-ingestion frameworks. Jay is a recognized thought leader and keynote speaker, guiding organizations in balancing innovation, trust, and accountability within security-critical and cloud-driven ecosystems. Jay holds an M.S. in Electrical Engineering from Cleveland State University and is focused on building secure, resilient cloud systems.