STO-Activities: (no title)

Activity title: RF based detection and classification of UAS
Activity Reference: IST-SET-204
Panel: IST
Security Classification: NATO UNCLASSIFIED
Status: Active
Activity type: RTG
Start date: 2023-10-03T00:00:00Z
Actual End date: 2026-10-03T00:00:00Z
Keywords: CUAS, Machine learning, RF detection and classification
Background: Unmanned aerial systems (UAS) are imposing alarming threats to the military installations, the national security and private sectors. In recent years, the threat has become increasingly vivid due to the wide availability of low-cost drones and the war in Ukraine showed the impact low-cost commercial drones can have on the modern battlefield. Furthermore, from the Russian-Ukrainian war theatre technical information is being acquired and shared on novel and improvised deployment of COTS and MOTS UASs, showing a rapid development of tactics and capabilities from both sides. Moreover, several illicit incidents at security-sensitive places such as airports, national campaigns, international sports events and nuclear power plants have been recorded. As an answer to counter the rising UAS threat, several activities have been organized within the STO. These efforts are mainly focusing on the challenging task of the detection and classification of UAS. SET-183 focused on modelling the signatures of LSS (Low, Slow and Small) UAS. This results in requirements that have to be met by a C-UAS system with respect to the drone characteristics that constitute the UAS signature.
SCI-SET-353 works on the development of a common modelling and simulation framework for C-UAS. The objective of SET-260 is to create a dataset of EO/IR signatures of UAS in order to develop and test detection and tracking algorithms, whereas SET-245 and now SET-307 address the topic of advanced radar-based detection, tracking, classification and recognition of UAS, including swarms and formations. The UAS detection methods utilizing different sensor domains each have their limitations. Radar-based detection systems may fail to detect and localize a drone due to a small radar cross-section (RCS), weather and daylight conditions affect the detection capabilities of EO/IR systems and the presence of high noise challenges acoustic systems. To overcome those limitations, a hybrid C-UAS system combining the sensing domains and therefore using the strengths of each approach should be preferred. The detection of communication between a drone and its ground control station (GCS) in the RF domain is complementary to the aforementioned approaches. An RF detector enables a C-UAS system to detect, classify and track a UAS through passive sensing without the need for line-of-sight (LOS) conditions. It therefore adds important capabilities to a hybrid system, with RF sensors being cheaper than radar systems. In the long term, RF detection can be considered the first step necessary to conduct a cyber attack on a UAS, since it allows to gather information regarding the drone and the pilot and ultimately to use it to take-over the drone.
In December 2021, IST-ET-120 on ‘RF fingerprinting of drones’ started a study to assess the need and the feasibility for a joint effort. The many efforts on a national level proof the importance of this research topic but they lack a large database and benchmarking the detection and classification techniques they are working on. The outcome of the ET shows the interest to (1) create a common database of UAS RF signals, (2) cooperate to exchange knowledge on classical, AI-based and hybrid techniques for detection and classification and (3) create a framework for benchmarking the developed methods.
Objectives: The projected RTG aims at bringing together RF researchers among the NATO community to:
 
1. Create a UAS RF database by gathering own RF measurements of various types of drones. The database structure should be designed such that integration of open source data is possible. Furthermore, it should be dynamic in nature, i.e. accommodating new entries related to the fingerprint of RF signals belonging to unidentified UAS types.
2. Identify the promising RF detection and classification techniques based on spectral and temporal information by assessing classical (e.g. matched filters) and AI-based (e.g. feature-based) approaches. The findings will be utilized to possibly develop a new hybrid algorithm for improved performance. Signal processing should also enable the classification of unknown UASs (e.g. novelty detection and/or incremental learning) and the dynamic assessment of situational awareness.
3. Define standard test and evaluation datasets in various simulated RF environments (e.g. rural, urban and industrial environments), in order to assess the performance of detection and classification algorithms.
 
The database will be designed, developed and populated in several stages, with a combination of technical activities, both in the lab and in the field, possibly including a final demonstration of the database’s functionalities. The research into relevant detection and identification algorithms could be realized by organizing a competition comparable to a “Kaggle” challenge. A common training dataset is shared between members, each member/nation tests different approaches (classical and AI-based) and submits the most promising ones. The submissions are tested against an unseen test dataset. This allows to gain insight where ML is a superior and more flexible approach.
 
The database, the algorithms and the knowledge acquired during the RTG will be shared amongst the partners and made available at broader NATO level.
Topics: 1. UASs RF emissions
2. RF detection and classification
3. Artificial Intelligence / Deep Learning
Contact:
Open2Partners:
Title:

Created at 10/07/2023 16:00 by System Account
Last modified at 16/05/2024 09:00 by System Account
 
Go back to list
Home(NATO STO)