Post by Seyed Hossein Ahmadpanah

🚀 New Paper Alert: "eBPF-Shield: Autonomous Anomaly Detection and Proactive Hybrid In-kernel Remediation for Cloud-Native Systems" is now published in Service Oriented Computing and Applications (Springer Nature). Cloud-native resilience is broken. Standard Kubernetes liveness/readiness probes only see binary "up/down" signals. They completely miss gray failures—CPU starvation, memory pressure, downstream latency, and I/O bottlenecks—that silently degrade services long before a pod crashes. We built eBPF-Shield to close that gap. It is a closed-loop, hybrid kernel/user-space framework that turns eBPF from a passive observability tool into an intelligent, autonomous shield: 🔍 Holistic, multi-modal sensing — We instrument the kernel's network, CPU scheduler, memory, and I/O subsystems simultaneously, giving every container a unique behavioral fingerprint. 🧠 Adaptive baselining — An lightweight EWMA model learns normal behavior per container in real-time, eliminating brittle, manually-tuned thresholds. ⚡ Proactive in-kernel remediation — A dynamically computed health score drives graduated load shedding directly in the kernel data path (TC/XDP), stopping cascading failures before they reach users. The results on a realistic microservices benchmark speak for themselves: ✅ Up to 90% reduction in time-to-detect for latency anomalies ✅ Detects complex gray failures completely missed by standard K8s probes and state-of-the-art network-centric eBPF baselines ✅ Maintains >85% success rate during injected faults (where baselines collapsed to <<40%) ✅ <<1% CPU overhead and sub-millisecond latency impact This shifts system reliability from reactive, user-space observation to proactive, kernel-aware self-healing. 📄 Read the paper: https://lnkd.in/dBQR9fRT Proud of this work with my co-authors Dr. M Mirabi and Sanaz Sobhanloo. Would love your thoughts, feedback, and questions! #eBPF #CloudNative #Kubernetes #SRE #AIOps #Microservices #Observability #LinuxKernel #SystemsResearch #AutonomousRemediation #GrayFailure #AcademicResearch