Classification:

attack

Tactic:

TA0001-initial-access

Technique:

T1078-valid-accounts

VIEW RULE IN DATADOG VIEW SIGNALS

Goal

Detects an Azure AI Foundry agent identity signing in with a client secret credential. Agent identities are expected to authenticate with certificates or federated credentials, not client secrets.

Strategy

This rule monitors Azure Active Directory service principal sign-in logs whose @properties.clientCredentialType is clientSecret. Agent identities backing hosted AI agents (for example Azure AI Foundry hosted agents) are designed to authenticate through certificate-based or federated credential flows, which do not expose a long-lived secret.

Triage & Response

• Identify the agent identity behind {{@ocsf.actor.user.uid}} and confirm whether it corresponds to a known, currently deployed hosted agent or AI workload.
• Review {{@properties.appDisplayName}} and {{@properties.resourceDisplayName}} to confirm the sign-in target matches the expected application or resource for this agent.
• Check the sign-in’s {{@network.client.ip}} and {{@network.client.geoip.country.name}} for consistency with where this agent identity is expected to run.
• If the activity is confirmed to be from a compromised agent identity, rotate or remove the client secet.
• Consider switching this agent identity authentication method to a certificate or federated credential if feasible.