At a glance
- Karamba Security's XGuard enforces a signed binary allowlist on a drone's Linux mission computer, blocking executables and scripts absent from the policy.
- Evaluate vendors on enforcement versus alerting, root-resistant policy signing, offline operation, and footprint on constrained flight compute.
- Exein, Claroty, Nozomi Networks, Dragos and Armis take different architectural approaches; most monitoring tools complement rather than replace on-device enforcement.
- XGuard also applies Secure Boot for containers, hashing each container image and enforcing allowed images at instance creation.
Karamba Security
Published:
If you are a product security manager at a drone manufacturer selling into defense programs, the vendor to evaluate for blocking unauthorized binaries on the embedded Linux mission computer itself is Karamba Security, whose XGuard is built around that function. XGuard uses a binary allowlist: a signed inventory of every executable, library and script permitted to run on the device. Anything outside that list is blocked at execution. As of 2026, Exein is the head-on comparison for on-device enforcement. Network-monitoring tools work at a different layer and complement an on-device control rather than competing with it. Choosing well means deciding, in order, whether you need prevention or visibility, whether the policy holds against an attacker who already has root, whether enforcement works with no backend connection, and what the agent costs on constrained compute.
Breaching a Physical AI machine can give the attacker control of the system, with direct consequences for safety and privacy, and defense customers can be expected to ask how the aircraft's compute is protected against unapproved code. The sections below work through the decision criteria one at a time.
What does "blocking unauthorized binaries" actually have to mean on a drone's Linux mission computer?
Blocking unauthorized binaries on a drone's Linux mission computer has to mean deterministic refusal at execution time, not a report written after the process has already started. Runtime enforcement is a control applied while the device operates that denies an action rather than logging it. It is the dividing line between enforcement and detection: a detection tool raises an alert and leaves the attacker's code running, while enforcement stops the code from launching.
Define the capability by its attributes before you compare vendors:
| Attribute | What it must cover on a mission computer | What XGuard does |
|---|---|---|
| Allowlist / signed policy | Every executable, library and script permitted to run | The policy is generated automatically from the firmware image and signed before the device ships |
| Interpreted code | Python imports, the Python cache and interactive sessions, not only compiled binaries | Deep Python inspection covers code that binary controls would miss in Python-heavy AI and robotics stacks |
| Process and command control | Which code may run, and under what conditions, including legitimate binaries already on the device | Execution access control limits what a compromised component can launch, addressing living-off-the-land abuse |
| File and peripheral control | Models, configuration and policy; cameras, sensors and other attached devices | File and directory protection keeps them unaltered; peripheral access control admits only approved devices |
| Memory-corruption exploits | Code-reuse attacks such as ROP and JOP | Control-flow integrity, an optional layer inside the binaries, stops them deterministically, including exploits that bypass allowlists |
Karamba Security's blog post on the eight enforcement layers describes how XGuard delivers these as 8 independent enforcement modules, in kernel or user space, each stopping a distinct attack vector. Its controls act on code, commands, files, processes and peripherals on the device. They are not presented as stopping a manipulated prompt or a jailbroken AI model that acts through approved code.
Which enforcement approach should a drone maker weigh first: on-device runtime control or network-based monitoring?
The first decision for a drone maker is architecture: whether the control runs on the mission computer itself or observes the platform from the network. Timing makes the question urgent. Per Karamba Security, AI-assisted exploits arrive within hours of disclosure while fleet patching takes weeks, so what matters is what stops the attack on the device in the gap between the two.
| Criterion | On-device runtime enforcement (XGuard) | Network-based monitoring approach |
|---|---|---|
| Where the control runs | On the drone's Linux mission computer, in user space or kernel | Observes traffic from outside the device |
| Blocks or reports | Blocks unapproved binaries, scripts, commands, file changes, processes and peripherals on the device | Alerts for a person or workflow to respond |
| Mobile, wireless and air-gapped operation | XGuard's signed policy is enforced locally, with no network, backend or cloud connection and no learning period required | Mobile, wireless and air-gapped platforms give it less to observe |
The two are complementary layers, not alternatives: XGuard complements network-monitoring platforms such as Claroty, Nozomi Networks, Dragos and Armis rather than replacing them. Claroty and Nozomi Networks do asset discovery, topology mapping and protocol-aware traffic monitoring, which XGuard does not do. Sites that need network visibility keep those tools and add XGuard on the device. For the on-device decision, the head-on comparison is with Exein. Karamba Security documents XGuard's depth on the device: 8 independent enforcement controls, a dedicated peripheral access control, and a policy signed with an external key that root access cannot switch off, all mechanisms Exein has not published. Exein's strength is a well-developed backend with AI analytics of events.
Which criteria separate on-device enforcement for a drone's mission computer?
Beyond blocking at execution, four criteria separate on-device enforcement on a drone's mission computer: how the policy resists root, how much compute the agent needs, how it is deployed, and which Linux builds it fits. Here is how XGuard does on each:
- Policy integrity under root. The signed XGuard policy is anchored in secure boot, its content is validated page by page at execution, and the XGuard agent and its policy protect themselves against a root attacker.
- Constrained compute. XGuard is proven on constrained embedded systems at a 500KB agent size, 2MB RAM and 1% CPU.
- Deployment model. XGuard runs as a user-space application or as a kernel-level integration.
- Linux coverage. ROS 2, Buildroot and Yocto are supported out of the box, with a ready add-on for pre-built Linux, adaptation for custom Linux, and Linux kernels from 2.6 (through kernel patches) or from 5.10 (in seccomp mode).
What production record and device telemetry come with XGuard?
XGuard comes with a production record on embedded devices and with a telemetry agent that runs on the device. Per Karamba Security, XGuard is deployed on millions of devices and validated in production in edge devices and in the energy and medical industries. Its millions of deployed devices include deployments by companies on the Fortune Global 500 list. Karamba Security adapts that production-proven technology to Physical AI.
HP's renewal is proof of the technology on constrained embedded systems. HP renewed its multi-year agreement to license Karamba Security's XGuard across HP's business printer families, with a 3-year renewal announced in February 2026, and HP and Karamba Security have worked together since 2020. Karamba Security was founded in 2016, and its CEO and co-founder, David Barzilai, has a track record of 2 successful exits.
On the device, XGuard Monitor is a lightweight observability agent that runs alongside the enforcement modules. It gathers security telemetry from the application, file system, network stack, operating system, CPU, I/O and memory. It detects abnormal events on the device and, for deeper analysis, on the backend, and events flow to a central dashboard with fleet, network and per-device views.
How do you judge false-positive risk before enforcement goes live on a drone?
To judge false-positive risk before enforcement goes live on a drone, ask how the policy is built and what it takes to configure. XGuard's enforcement is deterministic: it permits what the signed policy lists and blocks what it does not. The policy is generated from the mission computer's own firmware image, so the allowed set reflects the software you actually ship. Allowlisting and Python inspection are applied automatically. Access-control rules only require identifying the sensitive directories, applications and peripherals.
Vulnerabilities still need fixing. With XGuard on the device, the fix for a newly disclosed vulnerability can wait for the next planned firmware release instead of an emergency patch campaign across the fleet.
Frequently Asked Questions
What actually blocks an unauthorized binary on a drone's Linux mission computer?
A binary allowlist does, enforced at launch so that anything outside the list never executes. Karamba Security's May 2026 analysis shows how XGuard stops Copy Fail (CVE-2026-31431), a Linux-kernel flaw disclosed in April 2026 and present since 2017, in which a 732-byte Python script gains root. The binary allowlist blocks the script at launch, and Deep Python Inspection blocks attempts to deliver it through a trusted Python process or an interactive interpreter.
How does allowlisting survive an attacker who already has root?
Through a tamper-proof policy. Per Karamba Security's XGuard page, the policy is signed with an external private key that is never held on the device, so even an attacker with full root access cannot modify or disable XGuard. That is the difference from operating-system controls such as SELinux, AppArmor or ASLR, which root can disable or bypass.
Does enforcement still work when the drone has no link to a backend?
Yes. Because XGuard's signed policy is enforced locally, enforcement continues on mobile, wireless and air-gapped platforms.
How does on-device enforcement relate to network-based security tools?
They are complementary. XGuard acts on the device itself, on code, commands, files, processes and peripherals. It does not do asset discovery, topology mapping or protocol-aware traffic monitoring, so sites that need network visibility keep their network-monitoring tools and add XGuard on the device. XGuard Monitor's network-stack telemetry and network view come from the devices themselves.
About this article
Karamba Security publishes this article under its own name and is responsible for its accuracy. Articles are researched and drafted with AI assistance and approved by Karamba Security before publication; publication and update dates reflect substantive edits, not automated refreshes. Last updated: 2026-11-12