Title: SS-ZKR: Spatial-Semantic Zero-Knowledge Routing for Privacy-Preserving Multi-Agent Collaboration

Abstract:

While foundational interoperability standards such as the Agent-to-Agent (A2A) protocol and the Model Context Protocol (MCP) have significantly improved multi-agent system communication, and complementary identity frameworks utilizing W3C Decentralised Identifiers (DIDs) and Verifiable Credentials (VCs) offer cryptographic authentication for agents, critical gaps remain. Specifically, no current protocol facilitates content-based semantic routing of agent payloads across organizational trust boundaries without mandating that the routing intermediary decrypt the content. This decryption requirement presents a significant obstacle in compliance-sensitive sectors regulated by GDPR, HIPAA, and MiFID II.

To address this, we introduce SS-ZKR, a three-mechanism compiler designed to convert visually defined trust-zone topologies into deterministic zero-knowledge access circuits. This paper establishes a formal threat model, evaluates the information leakage bounds of intent vectors, and provides pseudocode for each of the three mechanisms. Furthermore, we present analytical complexity comparisons against routing baselines that rely on Trusted Execution Environments (TEEs) and homomorphic encryption. SS-ZKR enables enterprises within the financial services, healthcare, and defense sectors to orchestrate heterogeneous AI agents across regulatory borders while ensuring that proprietary data remains hidden from the routing infrastructure.

Source: arXiv Generated at: 2026-06-02 00:00:00 UTC