Various Videos & Talks

A Deep of Cisco AI Network Analytics: using ML in Wifi for anomaly detection, trend analytics, comparative analytics with a LIVE demo.

Cisco AI Endpoint Analytics and Detecting Spoofing attacks with ML/AI

Beyond Protocols: Why AI is Networking's Overdue Paradigm Shift

Cisco AI Endpoint Analytics and Detecting Spoofing attacks with ML/AI

Abstract: for the past decade, Artificial Intelligence (AI) and Machine Learning (ML) have been applied to networking in narrow, isolated ways—mostly anomaly detection, traffic forecasting, failure prediction—but the impact remains surprisingly and unfortunately too marginal. Core operations like troubleshooting, root-cause-analysis, and network optimization still depend on outdated, manual methods. Meanwhile, Generative AI (GenAI) and large language models (LLMs) are reshaping entire industries with agentic systems, tool learning, reasoning models, and real-time decision-making. Networking is lagging behind. Today’s uses—chatbots and config assistants—barely scratch the surface of what’s possible. This paper argues that the biggest opportunity in networking today is not another protocol or some algorithmic optimization—but a shift to intelligent, distributed, AI-driven systems. GenAI can unlock self-healing fabrics, predictive diagnostics, and cross-layer optimization at global scale. But the gap between potential and reality is widening fast. Now is the moment to act. The future of the Internet will be built by networks that learn, reason, and adapt. Those that don’t will be left behind.

Time to Revisit the Internet Layering Principle: AI-Driven Cross-Layer Optimization

Cisco AI Endpoint Analytics and Detecting Spoofing attacks with ML/AI

Abstract

The foundational principle of layered architecture, a cornerstone of the Internet's success, now

creates a performance barrier for high-value, modern applications due to its strict information isolation. Each layer operates with an architectural blind spot, optimizing its behavior using only the information available within its own domain. This paper argues that the time has come to augment this model with AI-driven cross-layer intelligence. By training machine learning models on holistic, multi-layer telemetry—including direct user experience feedback—systems can move beyond optimizing isolated network metrics. The new paradigm is to learn and predictively optimize for the high-level Service-Level Objectives (SLOs) that truly define performance, such as Quality of Experience (QoE) and Job Completion Time (JCT). This approach does not replace the layering principle, which remains essential for many tasks, but complements it through a non-disruptive

"soft layering" model. It represents a practical and necessary evolution for engineering the next generation of adaptive, high-performance systems.