ZT-COMM: Secure Agent-to-Agent Communication using Context-Aware Policies

Oct 12, 2026·
Amit Aneja
Amit Aneja
· 0 min read
Abstract
Two AI agents can complete a valid mutual TLS handshake and still be talking to an attacker. ZT-COMM adds mutual authentication, context-aware policy binding, continuous per-message integrity checks and a hash-chained audit log to agent-to-agent channels. In an open-source evaluation, it stopped all five tested attack types, including context manipulation, mid-session hijack and token replay, which mTLS-only, OAuth and allowlist baselines let through.
Type
Publication
Under review
publications
AI Agent Security Zero Trust Network Security
Amit Aneja
Authors
Amit Aneja
Senior Director, Product Management, Zero Trust Networking
Amit Aneja is Senior Director of Product Management at Zscaler, where he leads the Zero Trust Branch portfolio. A trusted technology executive with over two decades of experience in cloud infrastructure, hybrid cloud, and networking, Amit has a proven record of building and scaling large-scale platforms and $100M+ ARR products. At Microsoft, he led product strategy, OKRs, and multi-year roadmaps for Azure VMware Solution and Nutanix Cloud Clusters on Azure. His earlier career includes senior product roles at VMware (NSX, VeloCloud) and Cisco, giving him deep expertise across the enterprise networking and security stack. He holds a CCIE (Emeritus) in Routing and Switching and Service Provider.