STO-Activities: (no title)
Activity title:
CCDs Technologies to counter Artificial Intelligence Systems
Activity Reference:
SCI-359
Panel:
SCI
Security Classification:
Other
Status:
Active
Activity type:
RTG
Start date:
2023-10-17T00:00:00Z
Actual End date:
2026-10-18T00:00:00Z
Keywords:
Adversarial AI, Artificial Intelligence, Autonomous systems, Camouflage, CCD, COTS, Countermeasure, Dismounted Assets, Infrared, Object Detection, opensource, Signatures, Visible
Background:
The ISR (Information Surveillance and Reconnaissance) sensing capabilities of Armed Forces around the World has increased tremendously in the past few decades. Active and passive sensors operating in almost all spectral bands (where the atmosphere allows transmission) are being used at the strategic as well as at the tactical level. Some advanced sensors sensitive to polarization are even exploiting individual thin slices of spectral energy (hyperspectral) to extract detailed information about target signatures. In parallel, a new generation of autonomous and semi-autonomous platforms, such as unmanned autonomous ground (UGV) and airborne vehicles (UAV), is a growing and emerging threat to NATO forces in all theatre of operations.
So far, the amount of information generated by all different new sensors has overwhelmed even the best C4ISR (Computer, Communication, Command and Control, Information Surveillance and Reconnaissance) systems currently available. With this drawback, dismounted military assets with a less optimal camouflage and deceptive means could evade potential detection by a state-of-the-art reconnaissance system on the battlefield. However, new technologies based on Artificial Intelligence (AI) and Machine Learning (ML) software are currently emerging which shorten the information processing and decision-action making process. These game changer civilian technologies – which are often from commercial and/or open-source software codes and platforms – enable the processing of large amount of information from multiple simultaneous sources in near real-time. With no human assistance, camouflaged dismounted military assets could be tracked and detected by correlating multiple sources of information over different time and space domains, thus significantly affecting the effectiveness of Camouflage, Concealment & Deception (CCD) measures that have been proven to be effective so far.
In the context of the current Eastern European conflict, the use of AI-enabled technologies is the next logical step in the use of autonomous systems (for example UAVs/UGVs) for targeting dismounted military assets. Considering this, there is a paradigm shift in the doctrine in the use of CCD means. We need to assess the interactions between those new technologies and current CCD means in order to sustain a level of survivability on the modern battlefield. Previous NATO activities have been focusing on camouflage assessment from a conventional perspective with a human in-the-loop approach, ground-to-ground with human perception models. These recent NATO activities include SET-272, SCI-287, SCI 295, SCI-ET-039, SCI-325, and SCI-319.
Objectives:
The world of AI application is expanding rapidly especially in the open-source communities. This will certainly have serious repercussions in the defense community, especially in the ISR domain. Currently, most AI algorithms are easily accessible on the Internet and can be adapted and retrained to specific needs. It is also becoming easy to deploy them on miniature and low-cost computing platforms (i.e., edge computing) such as the one found on low-cost UAVs.
Thus, the main scientific objective of this project is to understand the interactions of open-source AI algorithms coupled with mobile sensors that could affect the performance of current and future CCD technologies (such as low observable dismounted military assets) in a close-range tactical engagement scenario.
The following scientific objectives will be considered in this activity:
• Determine the potential performance envelops of current CCD means against current and specialized open-source AI ISR algorithms.
• Assess how the COTS systems (sensors, platforms, algorithms) perform in operational setting against:
o Individual soldier, group of soldiers, snipers, soldiers with support equipment (i.e., logistic vehicles), etc.;
o Soldiers or group of soldiers with or without activities (static, walking, resupply, firing position, etc.);
o Decoy, camouflage nets, marker smokes, etc.;
o Individual thermal camouflage; and
o Facial alteration;
• Determine and assess the CCDs TTPs and their performance against the probable types of COTS autonomous systems with or without edge computing capabilities that could be threats for dismounted personnel. This includes:
o Environmental boundaries (extreme weather conditions) for deployment of autonomous systems;
o Current and near-term sensor capabilities payload on-board autonomous systems;
o Limits of communication, navigation, processing time, operating range, and altitude; and
o Potential swarming capability.
• Study methodologies (soft or hard, cyber, software, data) to counter open-source AI based targeting in operational conditions.
o Exploitation of loopholes in open-source AI algorithms (e.g., security) for deception and decoy purposes.
• Build a dataset oriented toward dismounted assets camouflage to share with NATO partners.
Topics:
This RTG will focus on the approach to be taken to understand the interaction of open-source AI algorithms and the role they could play in the ISR process against dismounted military assets. Here are some of the scientific topics to be covered:
• Effectiveness of CCD under various conditions against open-source AI driven ISR systems;
o Consideration of CCD variability to defeat AI;
o Effectiveness of camouflage/pattern/patches/fractal design under various observation ranges, illumination conditions, and viewing angles; and
o Effectiveness or design of adversarial patches (dual attribute human and AI, combined visible and thermal bands).
• Development of simple decoy against open-source AI algorithms (fire pits, BBQ, heated mannequin, etc.).
• Disruptive effect on dismounted assets operations.
• Development of standard testing methods and metrics against open-source AI algorithms.
o Description metrics between human and drones.
• Included in this evaluation will be an estimate of the current TRL and limitations of the technologies in military applications.
• Explainable AI applied to CCD means and dismounted military assets.
Contact:
Open2Partners:
Title:
Created at 15/05/2023 16:00 by System Account
Last modified at 16/05/2024 07:00 by System Account
Go back to list
Home(NATO STO)