Securing agent cards to prevent attacks before agent activation
We Have All the Cards: Securing Agent Cards at the Operational Trust Boundary of the Agent2Agent Protocol
Cryptography and SecurityArtificial Intelligence
Summary
Some computer programs called AI agents need to find and work with other agents by reading special cards with information. These cards use natural language, which bad actors can change to trick the system into choosing the wrong agent. The authors studied different ways these cards can be manipulated and created a security system called SECCARD that checks cards carefully to catch attacks. They tested SECCARD on millions of websites and many cards, showing it can spot nearly all attacks if the right language model judge is used.
What this means in practice
- •For security engineers: Detect and block manipulations in AI agent discovery metadata to improve security of multi-agent systems before invocation.
- •For cloud platform developers: Integrate SECCARD-like validation to secure agent interoperability frameworks used in cloud orchestration environments.
Authors
Miguel A. Guirao Aguilera, Satyajayant Misra, Efren Lopez-Morales
Abstract
AI agent interoperability protocols enable orchestrators to discover and invoke external agents using machine-readable metadata. In the Agent2Agent (A2A) protocol, Agent Cards contain attacker-controllable natural-language fields that may influence agent selection before a provider is invoked, creating a pre-invocation trust problem. We characterize this discovery-time attack surface through a taxonomy of ten Agent Card manipulation attacks and present SECCARD, a layered validation framework combining protocol checks, deterministic analysis, and LLM-based semantic reasoning. We measure the emerging public A2A ecosystem by probing 36.1 million Internet domains and collecting 310 deployed Agent Cards. Two annotators independently label the corpus with 96.8% agreement. We then construct a 660-card real-card perturbation benchmark with matched benign controls and a 260-card synthetic adversarial benchmark spanning all ten attack categories. On controlled attacks injected into real Agent Cards, the strongest semantic judge achieves 97.0% security recall at a 1.7% false-positive rate. Across nine LLM judges on the synthetic benchmark, security recall ranges from 25.8% to 92.3%, demonstrating substantial model-dependent variation. These results show that SECCARD effectively identifies Agent Card manipulations spanning all ten discovery-time attack classes while preserving benign Agent Cards, but also reveal that its effectiveness depends critically on the choice of the underlying LLM judge.