In the age of AI-powered hacking, keeping the U.S. military’s vital space-based capabilities functioning through enemy cyberattacks will require the Pentagon and its contractors to fully embrace autonomous on-orbit cyber defenses, a senior defense official told CyberSat this week.
“Adversaries are actively utilizing machine learning to discover zero-day vulnerabilities in satellite flight software, orchestrate autonomous multi-vector cyber attacks, and generate hyper-realistic synthetic telemetry to deceive our ground controllers,” warned Mike Dean, the director of Command, Control and Communications Infrastructure (C3I) for the Pentagon’s Chief Information Officer during a briefing on Nov. 3.
“To defeat this threat,” Dean added, “our defensive and offensive capabilities must harness the same frontier [artificial intelligence] technology.”
Infrastructure is the sine qua non of AI, Dean said, “AI cannot deliver military value if it sits on fragile, fragmented infrastructure.”
Algorithms are only as effective as the compute platforms hosting them or the data used to feed and train them, he explained. What’s required is an “enduring digital foundation, a unified, hybrid, multi-cloud environment that extends from the enterprise data centers out to the tactical edge and into space.”
Dean explained the Pentagon is building compute architectures that are capable of running containerized microservices and AI-driven, automated cyber defense directly onboard satellite payloads. This will provide a foundation for standardized data architectures and data curation, and for “hardened optical and RF transport networks capable of withstanding hostile electronic and cyber disruption.”
Speed is one of the defining characteristics of AI-assisted cyberattacks, so the Defense Department (DoD) has to get rid of outdated manual verification and code audits, and slow change control boards. They are being replaced with “continuous automated DevSecOps pipelines across the entire DoD enterprise,” Dean said.
Patching and other updates have to be possible within hours not months, he added.
“When an orbital vulnerability is discovered, or an adversary attempts novel electronic spoofing against our satcom constellations, we can’t wait months for human review boards or for the next physical satellite block build,” Dean said. “By automating testing against high-fidelity digital twins and streamlining automated signing, we can compile, validate, and beam a cryptographically verified patch directly to orbit within hours.”
Cyber defenses must be able to operate autonomously, because of the hostile character of the warfighting environment in which connectivity can be limited, interrupted or totally denied, Dean said.
“Commercial cybersecurity tools built for high bandwidth corporate environments fail at the tactical edge,” he said. Warfighting cyber resilience requires “deploying autonomous, lightweight AI agents capable of detecting anomalies, containing intrusions, and self-healing systems locally on deployed platforms without needing continuous connection to an enterprise cloud.”
If a satellite loses communication with the ground station during a cyber attack on-board defensive AI must be able to isolate the compromised subsystems, reroute telemetry, and preserve mission capability autonomously.
Resilience must be built in from the design stage, he added. Before the fight computer is integrated into a launch vehicle “it must undergo realistic, unconstrained adversarial testing of digital twins and hardware in the loop, software in the loop, cyber ranges.” The aim is to subject the military’s satellite software “to the most sophisticated nation-state exploitation toolset we have, and then see the holes.”
To ensure contractors keep up, the CIO’s office is “mandating the formal validation of cyber resilience requirements across all national security space systems, weapon systems, and interconnected platforms. Zero exceptions for space platforms,” Dean said.
More from CyberSat 2026:
——————————————————-
Click Here For The Original Source.
