DoD C3I director pushes autonomous cyber defenses for military satellites

Mike Dean, director of Command, Control and Communications Infrastructure for the Pentagon’s Chief Information Officer, said on Nov. 3 at a CyberSat briefing that contractors and the Defense Department must put fully autonomous cyber defenses on orbit, so orbital military operations can survive enemy cyberattacks, presenting the shift as necessary to keep satellite-based military operations running while adversaries use AI against orbital systems.

Via Satellite reported that adversaries actively use machine learning to find zero-day flaws in satellite flight software, launch autonomous multi-vector cyber attacks, and create hyper-realistic synthetic telemetry meant to trick ground controllers, based on Dean’s Nov. 3 briefing. He called the tactics current adversary behavior and said U.S. defensive systems must match the same frontier AI capabilities.

Dean said both defense and offense must rely on the same advanced AI technology that adversaries use to counter AI-powered hacking, because older tools cannot stop AI-driven attacks. He made the point immediately after describing adversarial machine learning uses, arguing that U.S. systems must match the frontier capabilities adversaries already use against orbital software.

Dean said autonomous operation is required for cyber defenses since warfighting connectivity may be limited, interrupted, or fully denied. A spacecraft that loses contact with a ground station during a cyber attack needs AI-based defense onboard to separate compromised subsystems, redirect telemetry, and keep the mission going without help from controllers when communication is cut.

The Defense Department is constructing compute architectures that run containerized microservices—small, independent services packaged in containers—and AI-based automated cyber defense aboard satellite payloads, placing protection close to the systems it guards. Dean said the capability would support standardized data architectures and data curation, along with hardened optical and RF transport networks that survive hostile electronic and cyber disruption.

Dean said updates and patches must occur within hours instead of months, using cryptographically verified patches that automated systems compile, validate, and send to orbit within hours through automated testing against high-fidelity digital twins—virtual replicas used for testing—and streamlined automated signing. He said the military cannot hold out for human review boards or the next satellite block build when an orbital vulnerability appears or an adversary attempts novel electronic spoofing against satcom constellations.

The Defense Department is swapping outdated manual verification, code audits, and slow change control boards for continuous automated DevSecOps pipelines—automated tools that build, test, and deploy software securely—since AI-assisted cyberattacks move quickly. Dean said AI-assisted attacks move at machine speed, and that review conducted at human speed is too slow to defend orbital systems.

The CIO’s office mandates formal validation for weapon systems, interconnected platforms, and national security space systems under its cyber resilience requirements, granting no exceptions to space platforms. Dean said the rule makes sure contractors keep up with Pentagon cyber resilience standards for space assets.

Dean said cyber resilience has to be built in from the design stage, and that pre-launch computers must undergo realistic unconstrained adversarial testing using digital twins, hardware and software in the loop, and cyber ranges. Pre-launch testing works to expose weaknesses by running military satellite software against the most advanced nation-state exploitation toolset the Pentagon has.

Commercial cybersecurity tools built for corporate networks fail at the tactical edge, Dean said, and warfighting resilience requires autonomous lightweight AI agents on deployed platforms to detect anomalies, contain intrusions, and self-heal systems without a continuous enterprise cloud connection. He contrasted corporate security tools with what deployed military platforms need under denied or intermittent connectivity.

Dean said AI offers no military value on fragile, fragmented infrastructure, and that an enduring digital foundation reaching from enterprise data centers to the tactical edge and into space is required. The effectiveness of algorithms relies on the compute platforms that run them and on the data used to feed and train them. The required foundation is a unified, hybrid, multi-cloud environment that runs from enterprise data centers to the tactical edge and into space.

Dean offered a direct assessment of the limits of AI on orbital hardware. “AI cannot deliver military value if it sits on fragile, fragmented infrastructure.” He said algorithms perform only as well as the platforms and data behind them, so the Pentagon needs a unified digital environment extending from data centers into orbit.