Defence
AI-enabled Electronic Warfare
Drone detection is an escalating problem
As unmanned systems continue to proliferate in modern warfare, traditional radar technology is struggling to keep pace.
In the case of UAVs, small low-cost drones can evade radar using sophisticated frequency-hopping techniques. The variety of platforms - from hobbyist quadcopters to military-grade systems - complicates identification. As do congested radio frequency (RF) environments filled with WiFi, Bluetooth, and other signals.
When new threats emerge, legacy counter RF solutions rely on pre-defined signature libraries that need lengthy updates. And while adversaries can modify their platforms in days, defenders are stuck with month-long update cycles, leaving them perpetually behind.
Why current solutions fall short
Physics-based, traditional signal processing detection methods struggle in dense, noisy environments. They lose critical information and miss subtle patterns that distinguish friend from foe.
Hardware-centric solutions face their own challenges. Most systems require expensive, specialized equipment that cannot adapt to new frequencies or protocols, making tactical deployment complex and prohibitively costly.
The challenge demands a fundamental shift: from hardware-defined to software-defined solutions. From reactive updates to adaptive learning. From specialized equipment to scalable, edge-deployable capabilities.
Faculty’s approach
At Faculty, we have developed a novel RF detection approach with machine learning at its core. As Europe's largest applied AI company, we've developed a software-first solution that retools military drone threat detection for the modern age.
Our system leverages raw IQ data directly, avoiding traditional pre-processing information loss. Using advanced transformer models, we detect the patterns that remain invisible to conventional systems and human operators, enabling drone signal classification even in congested RF environments. But it’s not just about UAVs, our methods aren’t rigid nor data specific. We have applied these approaches across different platforms and RF signatures.
MACHINE LEARNING ELECTRONIC WARFARE
We've compressed the threat response cycle from months to hours. Our end-to-end pipeline can go from recording a new drone signal in the field to deploying a trained ML model in just 24 hours. When adversaries evolve, we evolve faster.
We've compressed the threat response cycle from months to hours. Our end-to-end pipeline can go from recording a new drone signal in the field to deploying a trained ML model in just 24 hours. When adversaries evolve, we evolve faster.
Our models can run on commercial off-the-shelf hardware costing less than £1,000. We've proven deployment on devices like the NVIDIA Jetson Nano, bringing sophisticated EW capability to forward positions without requiring expensive, specialized equipment.
We've invested significantly in collecting real-world RF data from over 15 distinct drone platforms across varied environments – from anechoic chambers to urban battlefields. This operationally-relevant dataset, gathered through trials like Wintermute, with NPSA, or in international exercises, ensures our models perform in actual combat conditions, not just laboratories.
As a software solution, our system deploys across land, air, and maritime domains on existing hardware. We're not selling another box – we're providing capability that integrates with your existing infrastructure.
Our ensemble ML approach identifies specific drone platforms by their unique RF ‘fingerprints’, immediately determining friend from foe. The system continuously learns, pushing updates to edge devices as new threats emerge.
Faculty's technology addresses and solves drone detection threats: from reactive to predictive, from hardware-constrained to software-defined, from expensive to scalable. In accelerating electronic warfare, only adaptive AI maintains tactical advantage.