Counter-UAS Buyer's Guide 2026: How to Evaluate Drone Detection and Defence

Why are so many organisations buying counter-UAS now?

Two things changed at once: the money and the threat.

MarketsandMarkets' July 2026 release names two drivers. The first is coordinated drone swarms, which push demand for systems that can detect, track and mitigate several drones at the same time. The second is government funding for buying, testing and deploying counter-UAS. The same release expects mitigation and neutralisation to be the fastest-growing segment.

The threat also got harder to sense. Some drones no longer use a radio link at all. In Ukraine, fibre-optic-controlled FPV drones are flown over a physical cable, and one commander described the result as operating "in total radio silence" (The War Zone). A drone that flies a pre-programmed route can do the same. The US Cybersecurity and Infrastructure Security Agency (CISA) puts the consequence plainly in its November 2025 guidance: "Non-RF emitting UAS generally cannot be detected by RF systems."

For a buyer, this means the question is no longer only "how far can it see?" It is "what kinds of drone can it see at all, and how many at once?"

What does each detection layer see, and what does it miss?

CISA groups drone detection sensors into four types: radio frequency (RF), radar, electro-optical/infrared (EO/IR) and acoustic. It also covers Remote ID, the broadcast that compliant drones transmit. Each one senses a different physical signal, so each one fails in a different way.

Strengths and limits summarised from CISA's November 2025 guidance. The "Emits?" column describes the sensing technique.

CISA's recommendation follows from the table. It advises a "system of systems" that combines two or more sensor types, with sensor fusion to reduce detection errors.

Inside the optical layer: staring, scanning or PTZ?

"EO/IR" covers very different designs. How the camera covers the sky matters as much as the sensor inside it.

The time between looks matters most for fast, manoeuvring drones and for swarms. A tracker has to decide which new detection belongs to which existing track, and the longer the gap, the harder that gets. We worked through the arithmetic in Blink and It's Over. The common design that works well pairs fixed sensors for detection with a cued zoom camera for identification.

Cooled MWIR or uncooled LWIR?

Thermal cameras come in two broad classes. Cooled mid-wave infrared (MWIR) detectors are more sensitive and reach farther. Uncooled long-wave infrared (LWIR) cores are smaller, cheaper and use less power. In our own published assessment, cooled MWIR costs 10 to 50 times more, draws far more power, depends on a cryocooler with a limited service life, and is often export-controlled. If your site needs detection well beyond the range at which you can act on a small drone, a cooled sensor may be the right purchase. If you need many sensors around a perimeter, the cost and power of each one decide how many you can afford.

Detect, track, identify, mitigate: what is the difference, and what can you legally do?

Vendors often sell these four steps as one package. Buy them as separate decisions.

• Detect: something is in the air here.
• Track: it is moving this way, at this speed, and it is the same object we saw a second ago.
• Identify: it is a drone, not a bird or a cloud edge, and ideally which type.
• Mitigate: stop it, by jamming, spoofing, taking control, or physical means such as nets.

CISA draws the same line. It says detection technology and counter-UAS systems are often treated as the same thing, but they are not. Detection can be used on its own, with no mitigation capability.

The line matters because the law treats the two very differently, and the rules vary by country. The United States is a well-documented example. A 2020 interagency legal advisory from the Department of Justice, Department of Homeland Security, Federal Aviation Administration and Federal Communications Commission states that Congress authorised only the Departments of Defense, Energy, Justice and Homeland Security to carry out limited detection and mitigation against drones that threaten covered facilities. It warns that other bodies, including critical-infrastructure operators, stadiums and airports, may face federal criminal laws if they use mitigation. Jamming, spoofing, hacking, nets, projectiles and lasers are all named. The same advisory says that even some detection systems can raise surveillance-law questions, depending on how they work. CISA recommends consulting legal counsel before buying or deploying detection technology.

This is not legal advice, and laws change. The practical point for buyers is to ask your counsel what you may lawfully do at each site, before you write requirements. Many private operators end up buying detection, tracking and identification, then handing the result to the authority that is allowed to act.

What belongs on your counter-UAS evaluation checklist?

Use these criteria to score proposals side by side. Each one maps to a question in the next section.

1. Radio-silent drones. Can the system detect a drone that emits nothing: fibre-optic, pre-programmed or autonomous? If detection depends on RF, which other layer covers that gap?
2. Swarms and time on target. How many simultaneous tracks can it hold, and how often is each track updated? For moving sensors, how long is any bearing left unwatched?
3. False-alarm rate, and how it was measured. CISA warns that a high false-positive rate causes "alert fatigue." Ask for false alarms per day or per hour, at a named site, over a stated period, with birds, clouds and traffic in view.
4. Range claims tied to a target. A range figure means little without the drone size, the weather and the probability of detection behind it. Ask for detection range and stable-track range separately.
5. Cueing dependencies. Does any sensor need another sensor to tell it where to look? If so, what happens when the cueing sensor misses?
6. Emissions and licensing. Does the system transmit? CISA notes that in the US, radar is likely to need authorisation to transmit. CISA also notes that some detection systems emit radiation that can harm people or disrupt aircraft systems.
7. Open interfaces. Can it report into your existing command-and-control (C2) without a proprietary gateway? SAPIENT, published as BSI Flex 335, has been adopted by the UK Ministry of Defence as its standard for counter-UAS. CISA notes that siloed systems increase operator task saturation.
8. Export controls. Is the product or any key component subject to ITAR or other export controls? Get it in writing early. Export licensing affects schedule before it affects technology.
9. Total cost of ownership. Count sensors per site, masts, power, network, installation, subscriptions, maintenance and calibration. Moving parts such as gimbals, and cryocoolers in cooled cameras, are wear items.
10. Evidence an operator can review. When an alert fires, can the operator see the imagery that produced it? Can past detections be reviewed later, and reprocessed when the software improves?
11. Libraries and updates. If identification relies on a library of known drone signatures, how often is it updated, and does it work across manufacturers? CISA raises both points.
12. Cyber and supply chain. Can the vendor provide a software bill of materials (SBOM) and hardware bill of materials (HBOM)? Who can physically access the units?

What should you ask vendors in an RFP or pilot?

Put these questions in writing, and ask for the answers in writing.

• "Show us a detection of a drone with its radio link switched off. What sensor produced it?"
• "What is your false-alarm rate per 24 hours at a comparable site, and how did you count it?"
• "For your quoted range: which drone, what size, what weather, what probability of detection?"
• "How many targets can you track at once, and what is the update interval per track when that many are present?"
• "Which parts of the system transmit, and what authorisations will we need at our sites?"
• "Which C2 standards do you output today, and can we see a live integration?"
• "What is the export classification of each component, and what licences does our delivery need?"
• "What does year three cost, including support, subscriptions, spares and recalibration?"
• "Can our operators replay the imagery or data behind every alert from the last 30 days?"
• "Which detections did your system miss in your last field trial, and why?"

The last question is the most useful. A vendor that knows its failure modes has measured them.

How do you run a counter-UAS pilot that tells you something?

A demonstration shows that a system can work. A pilot shows whether it works at your site. Plan it like a test, not an event.

1. Write the threat and the scoring plan first. Name the drone types, approach routes and the area you must protect. Decide in advance how you will score detection, tracking and false alarms.
2. Measure the background before any flights. Run the system for several days with no test drones in the air. Every alert in that period is a false alarm or an unknown aircraft. This is your false-alarm baseline.
3. Fly cooperative drones with a ground-truth log. Record each flight's GPS log so every detection can be matched to where the drone actually was. Fly the drone types you named, including small ones, at several ranges, altitudes and headings, against open sky and against clutter such as treelines, roofs and cloud.
4. Include the hard cases. Fly at night and in poor weather if your threat model includes them. Where it is lawful and safe, fly a pre-programmed route with the drone's controller link off, and fly more than one drone at a time.
5. Score with numbers. Record the probability of detection against range, the time from launch to first alert, how long each track holds before it breaks, and false alarms per day.
6. Test the integration, not just the sensor. Send alerts into your C2 or video management system and have your own operators work them.
7. Keep the raw data. Ask the vendor to hand over logs and imagery from the pilot, so you can re-score them yourself or compare vendors on the same flights.

Check local aviation rules before flying test drones, especially near airports or restricted airspace.

Where does passive thermal staring fit?

This is our layer, so read this section with that in mind.

GOTEAM builds passive, AI-based thermal detection. Fixed uncooled thermal nodes image their whole field of view on every frame, and the detection runs on an NVIDIA Jetson edge unit at the sensor, so it keeps working offline. Because the sensor does not transmit, it needs no spectrum licence and cannot be jammed or geolocated by the target. It detects the aircraft by its heat and shape, so a fibre-optic or autonomous drone is visible in the same way as a radio-controlled one. Each confirmed track comes with an annotated video clip an operator can review, and tracks are reported over SAPIENT (BSI Flex 335 v2) to third-party C2. GOTEAM is detection and identification only: it does not jam, intercept or shoot down drones. It is ITAR-free; the uncooled core is EAR-controlled, and international engagements remain subject to Israeli export-control review.

The limits are real. We publish detection up to 400 m, and that figure is target- and scenario-dependent. Like any camera, a thermal node needs line of sight and is affected by weather. Covering 360° takes several fixed cameras. Range is estimated from the image, not measured the way radar measures it. That is why we recommend pairing staring thermal detection with RF and radar for reach, and with a cued zoom camera for identification.

If you want to test the detection layer before any meeting, you can upload your own thermal or EO clip to FlyBox, and a GOTEAM field device will return an annotated result.

Frequently asked questions

What is the difference between drone detection and counter-UAS?

Detection finds, tracks and identifies a drone. Counter-UAS in the full sense also includes mitigation: jamming, spoofing, taking control or physically stopping the drone. CISA treats them as distinct, and detection can be used on its own.

Which drone detection technology is best?

None on its own. RF, radar, EO/IR and acoustic sensors each miss different drones. CISA recommends a "system of systems" that combines two or more types with sensor fusion.

Can RF detectors find fibre-optic or autonomous drones?

Generally not. CISA states that non-RF emitting drones generally cannot be detected by RF systems. Radar, EO/IR and acoustic sensors can detect them, within their own limits.

Do I need a licence to run a counter-drone sensor?

It depends on the sensor and the country. A radar transmits, so it is likely to need authorisation to transmit; CISA says so for the US. Passive sensors such as thermal cameras do not transmit. RF detectors may raise surveillance-law questions. Ask your legal counsel about each site.

Can a private company jam or take down a drone?

In many places, no. In the United States, a 2020 interagency advisory says only specific federal departments have that authority, and that others may face federal criminal laws. Other countries have their own rules. Get legal advice before buying any mitigation capability.

How should I compare vendors' detection range claims?

Ask what drone, what size, what weather and what probability of detection each figure refers to, and whether it is first detection or a stable track. Then measure it in a pilot at your own site.

What false-alarm rate is acceptable?

That depends on how many alerts your operators can handle. Ask for false alarms per day at a comparable site, measured over days, not minutes. CISA warns that high error rates cause alert fatigue.

What is SAPIENT, and why does it matter?

SAPIENT is a sensor-to-C2 interface standard owned by the UK Ministry of Defence. Its interface control document is published as BSI Flex 335, and the MOD has adopted it as its counter-UAS standard. Choosing systems that support an open interface makes it easier to add or replace sensors later.

Is uncooled thermal good enough, or do I need cooled MWIR?

It depends on range. Cooled MWIR is more sensitive and sees farther, but costs more, uses more power and is often export-controlled. Uncooled LWIR suits dense perimeter coverage at shorter range. Match the sensor to the distance at which you can actually respond.

How long should a counter-UAS pilot run?

Long enough to measure background false alarms over several days with no flights, plus a planned set of cooperative test flights at different ranges, times of day and weather. A one-afternoon demo cannot show you either.

Independent commentary by GOTEAM. Not affiliated with or endorsed by any organisation named above. Nothing in this article is legal advice.