1-minute read
What a defender needs to know
Warlock ransomware used the legitimate Zoho Unattended Agent to maintain remote access, according to Cisco Talos. Talos asked eight researchers how defenders could interrupt an adversary’s next move; their answers include decoy accounts, behavioral credential-access alerts, RMM allowlists and limits on AI-agent sessions. Their guidance covers several attack paths in enterprise environments.
Make an approved remote-management roster, enforce it through application control or EDR, and investigate new agents, services and long-lived outbound sessions against support tickets. MITRE describes the service-and-egress sequence. Route decoy-mailbox activity to its own queue and test that an inbound message produces an alert. On Windows, look for abnormal access to lsass.exe followed by dump activity, rather than relying on a Mimikatz-only alert; MITRE’s LSASS entry describes that chain. Give each connected AI-agent run a short-lived identity and test revocation of one run at the egress control.
A decoy sees only interactions with itself, and an approved RMM product still needs session review. Verify the controls on your own systems: an unapproved remote agent triggers the intended block or alert, and a revoked agent run loses access while approved sessions keep working.