Modern AI systems rely on complex, high-performance networks to move vast amounts of data, synchronize distributed training, and serve models at scale. These networks form the backbone of AI infrastructure but, they also introduce a rapidly expanding attack surface. With adversaries leveraging AI to exploit protocol weaknesses and implementation flaws, securing the network fabric has never been more critical.
This session explores Secrets Management for AI Networks-an often overlooked but essential discipline for safeguarding access, ensuring confidentiality, and maintaining the integrity of network devices and systems. We will unpack how they are used within AI infrastructure and the common pitfalls lead to exposure. Attendees will gain insight into best practices for handling secrets securely across their lifecycle, and strategies to prevent, detect, and respond to breaches.
Guest Speaker:
Harsh Pathak is a Technical Program Manager at Meta, specializing in security for large-scale infrastructure systems. With experience spanning network security, platform resilience, and secrets management, he focuses on building secure foundations that enable AI and distributed computing at scale. Harsh combines a deep technical understanding with a practical approach to risk mitigation, helping bridge the gap between engineering and security strategy. He is passionate about advancing secure-by-design principles in the next-generation network.