1-minute read
What a defender needs to know
Microsoft Threat Intelligence observed agentic tooling find exposed application credentials on GitHub and use them for Entra service-principal sign-ins. One agent reached Azure Key Vault data exfiltration five minutes after searching for client secrets. A compromised vendor-owned multitenant application credential also reached multiple downstream customer tenants, so include connected vendor apps when scoping exposure.
Search service-principal and non-interactive sign-ins from October 28, or as far back as retained records allow, for 199.19.77[.]84 and 199.19.77[.]18. The latter is a Tor exit node that can carry unrelated traffic. For the same application, pair failures AADSTS7000215 and AADSTS7000222 with later successes, then follow app and tenant IDs into Key Vault, Storage, Azure Activity and Microsoft 365 audit records. Microsoft also observed operations from 65.87.223[.]26 and Azure AI reconnaissance from 157.22.126[.]128. IPs are starting points; app identity and resource operations show the reach.
Find leaked credentials in owned and other public GitHub repositories. Revoke or rotate exposed app credentials and replace secrets in any vault the app could read; deleting the leaked code leaves the credentials usable. Remove unexpected app credentials and narrow service-principal permissions. Already issued tokens remain valid until expiry. Verify that the old credential fails new sign-ins and reconcile resource access and token lifetime across affected tenants. Microsoft’s proposed workload-identity Conditional Access block needs Workload ID Premium and covers single-tenant apps registered in your tenant; coordinate with the vendor for multitenant exposure.