Counter-UAS (C-UAS) is the doctrine, technology, and tradecraft of detecting and defeating unauthorized or hostile unmanned aerial systems. It is not a single product or a single sensor. It is a kill chain that spans detection, classification, decision, and engagement - operating across the electromagnetic spectrum, the physical domain, and the human decision layer all at once. Done right, C-UAS looks boring: a drone shows up in the wrong place, the system sees it, classifies it, and the threat is neutralized before anyone notices. Done wrong, you end up with an airport shut down, a base surveilled with impunity, or a refinery on fire.
For quick-reference answers on related terms, systems, and authorities, see the Counter-UAS FAQ.
The threat
UAS fundamentally broke the cost curve for air attack. A commercial quadcopter costs a few hundred dollars. An FPV strike drone with a repurposed RPG warhead costs a few hundred more. A Shahed-class one-way attack drone costs in the low tens of thousands. A swarm of any of them can saturate air defense systems designed around manned aircraft at six, seven, or eight figures per intercept.
The threat is not theoretical. In Ukraine, small UAS account for the majority of tactical ISR and a significant share of precision strike. FPV drones costing under $500 routinely kill armored vehicles worth millions. Shahed and Geran-class OWA drones - propeller-driven Geran-2s, jet-powered Geran-3s, the emerging Geran-5 variants - cycle through Ukrainian skies in quantities that overwhelm even high-performance NATO air defense, with terrifying cumulative effect despite high per-engagement probability of kill. In the Middle East, Houthi forces have used modified commercial UAS and Iranian-designed OWAs against coalition vessels and commercial shipping. During Epic Fury, despite overwhelming US and Israeli air superiority, first-generation OWAs continued to penetrate some of the densest layered air defense on the planet. The intercept rate was near-total and the cost asymmetry was still unsustainable.
Domestically, the gap is wider. Unauthorized drone incursions over Barksdale AFB, Langley AFB, and Fort McNair have gone largely uncontested. A drone operator filmed and uploaded videos of flights over the Raven Rock Mountain Complex before anyone did anything about it. Airports in Denmark, Norway, and Germany have been forced to halt operations on bare drone sightings. Gatwick lost days to a ghost in 2018 and nothing operational has fundamentally changed since.
The asymmetry is the whole problem. The attacker needs a $500 drone and a YouTube tutorial. The defender needs detection, classification, tracking, decision authority, a legal basis to act, and an effector that doesn't set a neighborhood on fire - all coordinated in real time, all against a target with a radar cross-section smaller than a bird.
The Sense - Make Sense - Act continuum
Counter-UAS operates inside the Joint All-Domain Command and Control (JADC2) framework as a three-phase kill chain: Sense, Make Sense, and Act. Skipping any phase breaks the chain. Short-circuiting "Make Sense" in favor of raw sensor-to-shooter pipelines is how fratricide happens. Skipping "Act" authorities turns an expensive sensor network into an after-action report.
Sense
Detection is the first problem and it is harder than most buyers assume. Small UAS have tiny radar cross-sections (often well under 0.01 m²), fly at low altitude where ground clutter dominates, and can be trivially confused with birds, flying debris, or ground vehicles. No single sensor modality reliably detects every class of drone in every environment.
The sensor families are:
Cooperative RF - Receivers that pick up broadcast identification such as ASTM F3411 Remote ID, ASD-STAN prEN 4709-002 Direct Remote Identification (the FAA Part 89 family), ADS-B Out, and UAT on 978 / 1090 MHz. When the drone is compliant, you get position, operator location, and often airframe type. When the drone isn't compliant, you get nothing. Spoofing these broadcasts is trivial - cheap ESP32 boards can power a phantom swarm - which is why the data needs to be adjudicated against physics downstream.
SIGINT / ELINT - Non-cooperative RF: SDRs tuned across 900 MHz / 2.4 GHz / 5.0 / 5.8 GHz ISM bands, cellular bands, and the SHF bands used for custom or satcom-controlled platforms. Adds direction finding through beamforming, time-difference-of-arrival, or multilateration when you have enough geographically dispersed receivers. Fails completely against fiber-optic drones (Ukraine has normalized these), inertial-guided terminal strikes, and any platform that runs silent.
Radar - The uncrowned king of C-UAS against the airborne target. FMCW is the dominant modality against small RCS for simultaneous range and velocity. Phased arrays (AESA, MESA) enable electronic beam steering and track-while-scan. Mechanically scanned radars like the AN/MPQ-64 Sentinel (3-degree beam at ~30 RPM) give you roughly one paint every two seconds on a target - fine for cueing, not fine as your only source. The problem with radar is the same problem that makes drones hard: they're small and they look like birds. Ku-band and higher is where serious small-UAS radars live, and clutter rejection plus micro-Doppler classification are where the vendor differentiation happens.
Electro-Optical / Infrared - The industry pedantry obscures a simple fact: these are cameras. Multi-band suites typically combine visible, NIR (~700-1400 nm, low-light), SWIR (~1400-3000 nm, sees through fog and obscurants that wreck visible and NIR), MWIR (~3000-5000 nm, thermal from hot electronics and ESCs), and LWIR (~8000-14000 nm, the "people finder" band great for locating the operator). Uncooled LWIR dominates the man-portable market. Cooled MWIR is where high-performance long-range thermal tracking lives. EO/IR is your Positive Identification (PID) layer - gold standard for classification, lousy at wide-area search.
Acoustic - Phased microphone arrays with ML-based classification on propeller and ESC signatures. Range is measured in hundreds of meters to low single-digit kilometers in the best case. Outstanding corroborator and extremely hard to spoof with current adversary tradecraft. Degraded by wind, traffic, generator noise, and urban reverberation.
LiDAR - Emergent for C-UAS. Short-range (typically sub-kilometer for small UAS) but produces precise 3D point clouds. Interesting for perimeter security and terminal-zone classification. Cost-benefit is still worse than established modalities in most deployments.
UAS-as-a-Sensor - Your own drones as sensor platforms. Tethered for persistence, free-flying for coverage. Opens massive IFF and deconfliction requirements, but lets you reach angles and geometries your fixed sensors can't. Low-SWaP payloads like the EchoDyne EchoFlight or equivalent give you a mobile radar that didn't exist a decade ago.
Non-traditional: HUMINT and OSINT - Not a sensor in the electromagnetic sense, but in a force-protection context it may be the best pre-warning you'll get. LAANC database cross-reference, community tip lines, and social media posts from nosy or adversarial operators have all driven real interdictions.
Each of these has strengths, weaknesses, and specific ways it can be degraded or deceived. The Ultimate Guide to Counter-UAS Operations goes into each sensor class in depth.
Make Sense
Raw sensor data is not actionable intelligence. A radar track is not an engagement decision. An RF detection is not identification. "Make Sense" is the phase that turns fragmented, noisy, contradictory sensor reports into a unified operational picture an operator can actually act on.
This is where most "integrated" C-UAS solutions quietly fall apart. Putting four sensor feeds on one screen is a dashboard, not fusion. The operator is still mentally correlating the radar blip at azimuth 045 with the RF detection bearing northeast with the EO/IR track on a different monitor. That works fine in a clean environment with two objects. It collapses the moment a real incursion happens and the operator is drinking from 15 firehoses while the threat closes 500 meters.
Real Make Sense requires:
- Data normalization - Every sensor reports in its own coordinate system (WGS84, MGRS, local Cartesian), at its own update rate, with its own uncertainty expression (covariance matrix, CEP, nothing at all), across its own transport (serial, gRPC, MQTT/JSON, vendor SDKs, ASTERIX CAT-062 over UDP). Before a single track can be correlated, all of this has to reach a common spatiotemporal frame with a shared clock, normalized uncertainty, and preserved vendor metadata. CoT and Link 16 J-messages both have lossiness that drops meaningful identity context into unstructured fields or off the side of the table entirely.
- Sensor fusion - Multi-hypothesis tracking, optimal global assignment, identity provenance, and trust calibration across heterogeneous sensor classes. Fusion is a physics problem with mathematical solutions - not a job for a frontier LLM. The math is older than most of us and is load-bearing. See our Fusion & Decision Engine capability page and What is Sensor Fusion? for the deeper walk.
- Classification and threat assessment - Going from "there is a moving object" to "this is a DJI Mavic 3 with valid Remote ID loitering at 400 ft AGL 2 km from the perimeter" or "this is an unclassified fixed-wing on a direct intercept heading with no cooperative emissions." These are very different problems even when the kinematics look similar.
- Common Operational Picture (COP) - The human interface where Make Sense becomes decision. Fused tracks, sensor coverage overlays, confidence scoring, and recommended courses of action rendered at latencies measured in hundreds of milliseconds. Our COP is built so the absence of detections in a well-covered sector is visibly different from the absence of detections in an uncovered sector - because one of those is useful information and the other is dangerous ignorance.
Act
Act covers everything from doing nothing visibly (passive measures) through full kinetic engagement. The appropriate response depends on threat assessment, rules of engagement, legal authority, the operating environment, and the available effectors.
Passive measures - Camouflage (multi-spectral netting), dispersion, EMCON, and physical hardening. Unsexy, unbilled, and mandatory. Nobody is going to demo this at a trade show but every C-UAS program should have it in place before spending a dollar on active countermeasures.
Soft-kill: Electronic Attack - RF jamming (barrage, spot, sweep), GNSS denial and spoofing, and protocol-level exploitation. Doctrinally part of Electronic Attack under wider EW; sometimes called SKGBAD (Soft-Kill Ground Based Air Defense) to distinguish it from kinetic air defense. Indiscriminate in the RF domain - jam 2.4 GHz and you jam every Wi-Fi and Bluetooth device in the area. Federally illegal in the US absent specific statutory authority. Useless against fiber-optic or fully autonomous drones.
Soft-kill: Cyber effects - Targeting the UAS as a computing platform: firmware exploits, ground-control-station compromise, supply chain. Uniquely precise and attributable when it works. Perishable (patches happen) and requires extensive FISINT preparation. Does nothing if there's no exploitable control link in the first place.
Directed Energy Weapons (DEW) - High-Energy Lasers (HEL) for precision engagement of individual platforms via deflagration, component damage, or battery ignition - deep magazine, atmospheric-attenuation-sensitive, dwell-time dependent. High-Power Microwave (HPM) systems like Epirus Leonidas for counter-swarm via broad electromagnetic pulse - shorter range than HEL, wide cone of effect, and effective against unshielded electronics at scale. Both are effectively infinite-magazine as long as you have power, which is why they're the economically sustainable answer to saturating drone volumes.
Kinetic: projectile - Small arms (nearly useless against maneuvering SUAS despite being the first thing everyone reaches for), crew-served weapons, and purpose-built C-UAS gun systems firing programmable airburst munitions (Rheinmetall AHEAD and equivalents). What goes up comes down. Unsuitable for most CONUS environments and many OCONUS urban settings.
Kinetic: interceptors - Guided interceptor munitions (Coyote Block 2+ class), drone-on-drone net capture (Fortem DroneHunter, XTEND Scorpio with ParaZero DefendAir), and kinetic-impact interceptor drones (Anduril Anvil). Interceptor drones are compelling for CONUS and urban environments because they avoid fragmentation and UXO risk, and some allow forensic recovery of the target airframe for exploitation. Missile-class interceptors have favorable engagement geometry but brutal cost asymmetry against cheap threats.
Kinetic: explosive / fragmentation - C-RAM (Phalanx-class and land variants), proximity-fuzed air defense munitions. Highest single-shot probability of kill against individual high-value threats. Maximum collateral damage envelope. Reserved for last-resort and active combat scenarios.
Operator interdiction - Going after the human, not the drone. Often the most effective and most overlooked CONUS response. If Sense and Make Sense can localize the operator through RF direction finding, Remote ID operator broadcasts, or OSINT correlation, dispatching a response team to interdict ends the threat at its source - stops follow-on sorties, enables prosecution, avoids RF and kinetic collateral, and typically falls inside the legal authority of local law enforcement.
Emergent: acoustic counter-measures - Systems like Fractal's Acoustic Resonance Mitigation use focused sonic, ultrasonic, and subsonic waves to induce propeller turbulence and destabilize IMUs. Not operationally deployed at scale, but the physics is genuinely interesting and there are few obvious countermeasures.
Why layered defense matters
No single sensor detects everything. No single effector defeats everything. A layered C-UAS architecture has overlapping sensor coverage for redundancy and fusion quality, graduated effectors matched to threat characteristics and engagement geometry, and a fusion and decision layer that ties them together.
A competent layered architecture has:
- Long-range outer ring - Radar, high-gain SIGINT, and cooperative RF at maximum effective range. Classification confidence may be low; the job is early warning and buying time.
- Mid-range classification ring - RF fingerprinting, behavioral analysis, and multi-sensor cross-corroboration driving threat assessment.
- Close-range PID ring - EO/IR with laser rangefinder integration, acoustic corroboration, and (optionally) LiDAR for high-confidence identification in the terminal zone.
- Soft-kill layer - EW and cyber effects engaged at range against threats with exploitable control links.
- Hard-kill layer - DEW and kinetic effectors for threats that penetrate soft-kill or are immune to it (fiber-optic, autonomous, hardened).
The layers have to be connected. A radar detection has to cue the EO/IR turret. An RF classification has to feed the engagement decision. A soft-kill attempt that fails has to escalate automatically to the next authorized option inside the engagement timeline. That requires a platform that fuses across all sensor types and automates decision workflow with auditable policy - not a collection of vendor tools operated by separate teams on separate screens.
The DDIL and edge deployment problem
Most C-UAS systems are architected assuming reliable connectivity to a cloud or command center. In practice, the environments where C-UAS matters most - forward operating bases, critical infrastructure, mobile CPs, event venues, airfields during contingencies - often have degraded, contested, or complicated connectivity. Russian EW capability demonstrated in Ukraine (Krasukha-4, Murmansk-BN, Zhitel) can degrade theater-scale SATCOM, GPS, and tactical data links. Chinese A2/AD doctrine explicitly targets communications infrastructure. Cyber and information warfare can compromise communications at the human layer before a single antenna goes down.
A C-UAS platform that depends on cloud processing for fusion or policy enforcement will fail exactly when it is needed most. Edge-deployable C-UAS runs the full stack - sensor ingest, fusion, policy, decision support, and the operator interface - on hardware that deploys with the unit. Connectivity enriches the picture when it exists; its absence does not destroy it. See Why JADC2 Needs Sensor Fusion at the Edge for the longer argument on why this architectural choice determines whether your platform survives contact with reality.
Legal authority in CONUS
Under 10 USC 130i and related statutes, statutory authority to interdict UAS in the United States is restricted to DoD, DHS, DOJ, and DOE in most circumstances. State and local law enforcement, critical infrastructure operators, and private entities generally lack legal authority to jam, spoof, or kinetically engage UAS, regardless of how threatening the incursion. This is the single biggest gap in CONUS C-UAS posture. It is being addressed legislatively at a pace that does not match the threat. In the meantime, detection, HUMINT/OSINT-driven operator interdiction through law enforcement, and passive measures are what most non-federal operators have available.
Consult legal counsel before deploying any active countermeasure. This page is not legal advice and is not a license to start jamming.
Where Empyrean fits
Empyrean's CUAS and Force Protection capability was built around these principles from the beginning. Multi-sensor fusion correlates radar, RF, EO/IR, acoustic, cooperative broadcasts, ISR, and HUMINT/OSINT into unified fused tracks with identity provenance. The Fusion & Decision Engine runs multi-hypothesis tracking with optimal global assignment - no greedy matching, so the 25th detection in a dense scene gets the same association quality as the first. The EMSO workspace provides spectrum awareness of drone control channels and EA planning with physics-backed RF propagation. The Policy & Decision Layer enforces ROE and engagement rules locally with a full audit trail, whether connected or not. The Common Operational Picture is the operator surface where all of it comes together. And the Simulation & Wargaming engine runs the same fusion and policy stack against synthetic sensor data, so training transfers directly to operations.
The whole stack deploys at the edge. Same software everywhere. Same operator experience whether connected or dark.
Going deeper
For an 18,000-word walkthrough covering every sensor modality, deployment consideration, effector type, PACE planning, and legal authority in operational depth, see The Ultimate Guide to Counter-UAS Operations. For the architectural case on why fusion and decision automation have to run at the edge rather than the cloud, see Why JADC2 Needs Sensor Fusion at the Edge. For the mathematical foundation of multi-sensor fusion itself, see What is Sensor Fusion?.
