Ashok Subbabhatta Gopalakrishna · 2026 · arXiv
DOI: 10.48550/arXiv.2610.26501 · open-access PDF
Los agentes de la IA en los contenedores actúan abriendo las conexiones de red, y el texto que las conduce puede ser hostil. Los controles colocados entre el agente y sus herramientas son ignorados por un proceso infantil o un conector directo, y su aprobación no está vinculada a la conexión que el núcleo abre más tarde. Presentamos el escudo de conexión: cada conexión TCP de salida de un pod supervisado está aparcada dentro del camino connect() del núcleo, antes de que exista cualquier paquete, por un programa eBPF de instrucciones 98 en el cgroup del pod. Un interceptor de nodo atribuye el conector mantenido a su cookie de conector del núcleo, proceso, digesto ejecutable, cgroup y
Machine translation (Meta M2M100); the English original follows.
AI agents in containers act by opening network connections, and the text that drives them can be hostile. Controls placed between the agent and its tools are bypassed by a child process or a direct socket, and their approval is not tied to the connection the kernel later opens. We present connection escrow: every outbound TCP connection from a supervised pod is parked inside the kernel's connect() path, before any packet exists, by a 98-instruction eBPF program on the pod's cgroup. A node interceptor attributes the held socket to its kernel socket cookie, process, executable digest, cgroup and
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.26501, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange