This paper highlights the rapid adoption of agent AI based on Large-Scale Language Models (LLMs) that is transforming the enterprise ecosystem, where autonomous agents execute complex workflows. However, LLM-based multi-agent systems (MASes) suffer from several critical security vulnerabilities, including fragmented identity frameworks, insecure communication channels, and inadequate defenses against Byzantine agents or adversarial prompts. We demonstrate that existing security strategies cannot effectively address these risks and propose BlockA2A, the first unified multi-agent trust framework that enables secure, verifiable, and interoperable agent-to-agent trust. BlockA2A employs fine-grained cross-domain agent authentication using decentralized identifiers (DIDs), a blockchain-based ledger for immutable auditing, and smart contracts that dynamically enforce context-aware access control policies. Furthermore, we propose a Defense Orchestration Engine (DOE) that actively neutralizes attacks through real-time mechanisms such as Byzantine agent flagging, reactive execution suspension, and immediate privilege revocation. Experimental results demonstrate that BlockA2A and DOE operate with sub-second overhead, enabling scalable deployment in LLM-based MAS environments. We formalize how to integrate Google's A2A protocol into existing MASs through a practical implementation.
