A senior Pentagon official has told satellite builders that ground-based security operations cannot keep up with attacks on spacecraft, and that defense has to move on board.
Mike Dean runs command, control and communications infrastructure for the Defense Department’s chief information officer. Speaking at CyberSat on November 3, he described adversaries using machine learning to find zero-day flaws in flight software, coordinate multi-vector attacks and fabricate telemetry realistic enough to fool ground controllers.
Defense that fits on a payload
Dean’s answer is a compute architecture that runs containerized microservices and automated cyber defense directly on satellite payloads. Because an attack can sever the link to the ground, those defenses must work alone: lightweight AI agents that spot anomalies, contain intrusions and restore systems locally without reaching an enterprise cloud.
If a satellite loses contact during an attack, Dean said the onboard system has to isolate compromised subsystems, reroute telemetry and hold the mission together. Commercial tools built for high-bandwidth corporate networks, he argued, fail at the tactical edge.
Patching at orbital speed
The department is also retiring the manual review boards that gate code changes. Dean wants continuous automated pipelines, with testing against digital twins and automated signing, so a developer can compile, validate and transmit a cryptographically verified patch to orbit in hours rather than months.
His office is mandating formal validation of cyber resilience requirements across national security space systems, with no exception carved out for spacecraft. Before integration onto a launch vehicle, flight software must survive adversarial testing drawn from the government’s own nation-state exploitation toolset.
The shift changes what buyers will expect from suppliers: an architecture that can be attacked, recovered and updated while it flies, not a security package bolted on at delivery.
