The most valuable component inside a captured autonomous system may not be its sensor, propulsion system, or airframe. It may be the storage device holding the information that reveals how the platform operates.
In an adversary’s hands, mission routes, sensor collections, communications settings, system logs, software configurations, AI models, target-recognition data, navigation parameters, and operational limitations could expose far more than the details of a single mission. Any of which could help another nation understand how the system senses its environment, makes decisions, communicates, navigates, and responds under real-world conditions. That knowledge could accelerate efforts to replicate the technology, identify vulnerabilities, develop countermeasures, or defeat the capability altogether.
Iran’s reported recovery of a U.S. autonomous underwater vehicle near the Strait of Hormuz should bring this risk into sharper focus. U.S. Central Command has stated that the recovered vehicle was an older model that did not collect sensitive information or carry classified sonar or radar equipment. There is no indication that this particular incident resulted in the loss of classified data.
The more consequential question is what happens when the next recovered system does carry sensitive mission data or advanced autonomous capabilities.
Autonomous and unmanned platforms increasingly operate where continuous communications and physical recovery cannot be guaranteed. They must often collect, process, and store information locally because network access may be intermittent, degraded, denied, or intentionally restricted.
That makes onboard storage essential to the mission. It also makes storage a high-value target.
Once an adversary physically possesses a platform, traditional network defenses become largely irrelevant. The adversary can remove its storage, connect it to specialized equipment, copy encrypted sectors for extended analysis, examine the boot process, search for key material, and repeatedly attempt exploitation in a controlled laboratory environment.
If the stored information is insufficiently protected or requires no authentication to release the hardware encryption key, the consequences can extend beyond exposing classified files. The platform may contain software, AI model files, system configurations, sensor-processing parameters, maintenance logs, mission histories, or other data that reveals how a billion-dollar capability was designed and how it behaves operationally.
Data-at-rest protection therefore cannot be treated as an administrative feature added after the platform is built. It must be part of the system’s architecture & hardware design from the beginning.
It is tempting to answer this problem by saying, “The drive is encrypted.” But encryption alone is not enough. Encryption protects data only when the encryption key is also protected. If a captured platform automatically releases its key without adequately verifying that the system or operator is authorized, the presence of encryption will do little to stop an adversary from accessing the data.
The critical question is not simply whether the bits are encrypted, it is, what prevents an unauthorized person who possesses the entire system from causing the drive to unlock?
That is the role of authentication.
Cigent Pre-Boot Authentication requires authorization before the protected operating system can load and before readable access to the encrypted storage is granted. Authentication takes place outside the operating system, preventing an attacker from relying on the normal OS environment as the gatekeeper to the data.
An adversary cannot even copy the encrypted contents of the drive without authorization as those sectors remain unreadable. Physical possession of the storage device does not automatically become possession of the information stored on it when data at rest is properly configured.
For autonomous systems, that separation is fundamental.
When an autonomous platform carries classified information, a single encryption mechanism should not be the only barrier between an adversary and the data.
NSA’s Commercial Solutions for Classified Data-at-Rest Capability Package uses independent layers of encryption and authentication to protect classified information. A properly designed solution combines an outer hardware full-drive encryption layer with pre-boot authentication and an independent inner software encryption layer with its own authentication mechanism.
Cigent can provide both sides of this architecture:
This is more than simply encrypting the same data twice. Each layer must maintain its own cryptographic boundary and access controls so that the failure or compromise of one layer does not automatically expose the underlying information. For a captured autonomous platform, an adversary would have to defeat independent encryption and authentication mechanisms while working outside the trusted operational environment.
That is the level of protection the capture scenario demands.
Most systems are designed around how they will function while everything goes according to plan. Autonomous platforms must also be designed around what happens when the mission does not.
Those questions should drive requirements for encryption, authentication, key protection, lockout behavior, and secure sanitization. Cryptographic erasure and block-level sanitization may provide additional options when loss is detected or considered imminent, but they should complement, not replace, persistent encryption and strong authentication.
The objective should be straightforward: even if the United States loses the platform, it should not also lose the information required to understand, reproduce, or defeat the capability.
The recovered underwater vehicle reportedly did not contain classified or sensitive mission information. That may accurately describe this incident, but it is not a sufficient security standard for the next generation of autonomous systems.
Classification labels alone do not capture every consequence of loss. Individually unclassified files can become highly revealing when aggregated. Software configurations, AI models, sensor data, route histories, operational logs, and system parameters may collectively expose capabilities that took years and billions of dollars to develop.
DAR protection will not prevent an adversary from examining visible hardware or physically disassembling a captured vehicle. It can, however, deny access to the stored data and software that would make that examination dramatically faster and more valuable.
As unmanned aircraft, underwater vehicles, surface vessels, ground systems, tactical servers, and AI-enabled edge devices become more capable, some will inevitably malfunction, become stranded, or fall into adversary hands.
The security architecture must assume that physical control can be lost.
When that happens, the success of the mission should not depend on recovering the platform before the adversary reaches its storage, risking valuable resources and American lives. Hardware full-drive encryption, pre-boot authentication, independent software encryption, and secure sanitization (in both hardware and software) can ensure that capturing an autonomous system does not mean capturing the knowledge inside it.
Because encryption without authentication is not enough, and possession of the machine should never become possession of its mission.
Conner Crisafulli is a solutions engineer and cybersecurity professional at Cigent with a unique background bridging elite military operations and advanced data security. Before joining Cigent, Conner served six years as a U.S. Air Force Combat Controller, where he specialized in high-stakes mission planning, communication systems, and joint operations coordination/execution. Drawing from his experience in complex, contested environments, Conner now helps federal agencies and enterprise clients strengthen their data-at-rest protection strategies. At Cigent, he focuses on practical applications of self-encrypting drives (SEDs), pre-boot authentication (PBA), and various CSfC (commercial solutions for classified) technologies to safeguard sensitive data against evolving cyber threats.
