DNS-like Trust Layer for Secure, Scalable AI-Agent Deployments on Kubernetes | Akshay Mittal, PayPal

DNS-like Trust Layer for Secure, Scalable AI-Agent Deployments on Kubernetes | Akshay Mittal, PayPal

by Akshay Mittal

Applied Sciences & Engineering Computing & Cybersecurity UTComputer networking and communicationsUTNNetwork security
🎙 Akshay Mittal 👥 5K 📅 October 24, 2025 ⏱ 26 min 👁 141 📄 expert opinion 🧭 2026-08-15
Available in: English (current) Français

Keywords

Agent Name ServiceDNSKubernetestrustMLOps

Summary

Akshay Mittal from PayPal presents the Agent Name Service (ANS), a DNS-like protocol for secure and scalable AI-agent deployments on Kubernetes. He begins by highlighting the shift from human-supervised ML to autonomous agent ecosystems, which introduces trust and security challenges. ANS addresses these by providing a naming and discovery mechanism that maps agents to cryptographic identities, capabilities, and trust levels. The talk covers the protocol design, including a hierarchical naming convention and integration with existing standards like A2A, MCP, and ACP. A live demo showcases ANS registration, resolution, and verification with sub-100ms performance, along with Kubernetes-native deployment using CRDs, admission controllers, and service mesh. Performance metrics from the demo show deployment in 15 seconds, service creation in 5 seconds, and monitoring setup in 10 seconds. The speaker outlines a roadmap for full implementation, including advanced features like zero-knowledge proofs and multi-protocol support. He emphasizes the open-source nature and invites community contributions. The talk concludes with a Q&A session addressing service mesh integration and adoption strategies.

165 words

Critical Evaluation

Value of the Information & Strength of the Argument

The talk provides a valuable overview of a practical solution to a real problem in agentic AI: trust and governance. The speaker demonstrates a working proof-of-concept with performance metrics, which adds credibility. However, the argumentation is largely based on the speaker’s own design and implementation, with limited external validation. The claims about scalability and security are plausible but not rigorously proven. The speaker acknowledges the early stage of the project and outlines future work, which is honest but also indicates the lack of mature evidence.

Scientific Rigor, Source Quality, Title Accuracy

The talk is based on the speaker’s own research and implementation, with references to industry protocols like A2A, MCP, and ACP. The description provides a link to the MLOps World conference, but no direct sources are cited in the talk. The title accurately reflects the content. The speaker mentions using AI for documentation, but this is not a source. Overall, the scientific rigor is moderate, with a clear proof-of-concept but lacking peer-reviewed validation.

173 words

Title / Content Match

The title accurately reflects the content: a DNS-like trust layer for AI agents on Kubernetes, presented by Akshay Mittal from PayPal.

Quality & Reliability

6/10

The talk presents a proof-of-concept for a DNS-like trust layer for AI agents, with live demos and performance metrics. However, it lacks peer-reviewed validation, detailed security analysis, and independent verification. The speaker is a staff engineer at PayPal, lending practical credibility, but the claims are largely anecdotal and based on a small-scale pilot.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Concurring Sources

Contribution & Novelties

The talk introduces the Agent Name Service (ANS), a novel protocol for trust and governance in AI-agent deployments. It provides a concrete implementation on Kubernetes with performance benchmarks, which is a practical contribution to the field. The approach of combining DNS-like naming with cryptographic identity and capability verification is innovative.

Pour aller plus loin :

  • Agent Name Service GitHub repository — Note: The speaker mentioned a GitHub link, but it is not provided in the description. This is a placeholder; the actual repository is not verified.
  • Open Policy Agent (OPA) — Policy enforcement tool used in the talk.
  • Sigstore — Software signing and verification, mentioned for supply chain security.
  • A2A protocol — Agent-to-agent protocol by Google, referenced in the talk.
  • MCP protocol — Model Context Protocol by Anthropic, referenced in the talk.

132 words

Radar Profile

The radar profile shows a balanced distribution across quantity, quality, and technical level, with a slightly lower reliability score. This indicates a technically informative talk with practical demonstrations, but with limited external validation and scientific rigor.

Reliability 5/10