A dead drop resolver is a command-and-control technique that hides the address of the real C2 server inside content on a legitimate, trusted third-party service. Malware reads that content, a profile bio, a paste, or a dataset, to learn where to connect, so there is no hardcoded C2 domain to block and the traffic to a well-known service looks normal. In agentic intrusions, attacker-controlled datasets on a compromised platform act as the dead drop.
What a dead drop resolver is
A dead drop resolver decouples the malware from its C2 address. Rather than shipping a hardcoded domain that defenders can block, the malware fetches a pointer to the current C2 from content hosted on a service the network already trusts. The attacker updates that pointer at will, and the malware always finds its way home.
How it works and example
The attacker plants the C2 location, often encoded, in an ordinary-looking place: a social profile field, a paste, a repository file, or a dataset. The implant reads it and connects, so its only visible traffic is to a reputable service. In a 2026 agentic intrusion, attacker-controlled datasets on the compromised platform served as asynchronous message queues, a dead drop for commands and results.
How to defend against it
Domain reputation will not help, so watch behaviour: flag programmatic or repeated access to content services from server workloads, restrict which hosts may reach them, and inspect fetched content for encoded pointers. Combine egress allowlisting with anomaly detection on timing and volume, and treat unexpected reads of public content from production systems as suspicious.
References
- [1]MITRE ATT&CK: Command and Control (TA0011)(MITRE ATT&CK)
- [2]HuggingFace, Agent intrusion technical timeline(HuggingFace)
- [3]MITRE ATT&CK: Web Service (T1102)(MITRE ATT&CK)
If your workloads can read arbitrary public content, they can find a hidden C2. Talk to a security expert about egress control and detection.