DGA Engineer Creates “Anti-IED” Artificial Intelligence – FOB – Forces Operations Blog

What if a simple smartphone with a layer of artificial intelligence was enough to counter an improvised explosive device (IED or IED)? This is the question answered by a young engineer from the French Armaments Directorate, who created a solution first presented at the Defense Innovation Forum (FID).

IEDs that are difficult to detect

Behind the acronym “Detection of embedded link by interception of electromagnetic communication” (DELICE), a concept study led by engineer Axel, in charge of the experience in artificial intelligence within the DGA information control (DGA MI). Joined the Ministry of the Armed Forces in 2019, he committed to responding to a need that arises from OPEX theaters: to go one step ahead by instantly detecting and identifying malicious communications linked to IEDs.

The signals at the source of the triggering of an IED are “very difficult to detect at the moment,” acknowledges this deep learning expert, because they are emitted in specific bands of the electromagnetic spectrum by transmitters that can be acquired for three times nothing on AliExpress. “It is really a need expressed at the program level, based on a problem that has arisen on the ground. There is an unknown component that I cannot measure, identify, so I need a reliable solution ”, he explains.

The danger is daily for the operation, which faces a jumble of friendly and potentially enemy firms that are particularly difficult to unravel. Therefore, DELICE’s main objective was to achieve “to achieve this commitment to discriminate and encode the threat while keeping our own communications intact.”

Based on this observation, “we are proposing a deep learning model that will accurately identify the electromagnetic parameters of this threat.” To achieve this, “we have used several thousand samples of the threat that we want to code and a few thousand samples of what we do not want to code at all.” This library was then used to train a deep learning model based on the DGA MI computing infrastructures, “big bays full of graphics processors.” [GPU] ».

Towards an AI on board

Once in working order, this model was converted, optimized and integrated into a mobile phone. Why this type of platform? “Because we wanted a proof of concept that could be on board. The work context is a truck within a logistics convoy. So we need something that is transportable ”. A choice that goes against the image conveyed by AI, often illustrated by huge rooms lined with several thousand GPUs.

“What is true for learning is not true at all for use”, underlines the engineer Axel, who specifies that “we are now perfectly capable of developing models that run on the phone.” Connected to a signal acquisition kit, a smartphone would be sufficient to continuously provide a communications interception capability in the band in which the threat is located. Once the event is identified, it will be analyzed, characterized and represented by an intuitive color code: blue for friendly communications, green in the absence of risk, orange and red for the opposite situation. In the case of an IED, DELICE will automatically activate zone interference.

The engineer Axel behind DELICE, presented at the Defense Innovation Forum (Credits: Ministry of the Armed Forces)

“We have developed a sufficiently rapid alert system. That’s what deep learning is used for: generating a protective field even before the end of the malicious message ”. Because it is detected early enough, this threat will only require triggering interference in a relatively short period of time, limiting potential interference with friendly communications.

DELICE also navigates the richness of the electromagnetic spectrum to add a string to your bow: the fight against drones. “Just as we train our algorithm to detect an IED threat, we train it to identify different drone remotes based on their communication protocols.” Supporting evidence with a successful demonstration based on class 1 and 3 drone remotes.

Although complex, the creation of the anti-IED function was completed in record time. Between collecting the data, processing it, and implementing the model on the phone, it only took a month and a half. Thanks to these achievements, repeating the process for the anti-drone function only took about ten days. Ultimately, it is industrial integration that will require patience, on the order of several years. A “completely normal timing in view of what is at stake”, estimates the engineer of the DGA.

“Muscular” the BARAGE kit

This AI alone will not be able to guarantee complete protection against IEDs, many of which are activated by other means, such as a pressure system. However, it will help jump-start a program that is already underway, BARAGE. Designed by Thales, this kit creates a blocking bubble around a disassembled vehicle and fighters. In particular, it will be installed on PPT trucks. For the engineer Axel, “DELICE brings all the necessary elements to prepare future increases and make our jams more receptive and more effective against new threats”.

We really are on the premises of an AI as close as possible to the sensor, which works autonomously, with the media at hand and without WiFi or 5G ”. If the anti-IED function is indisputably linked to BARAGE, the anti-drone blind would be closer to the semi-fixed solutions designed for the protection of sensitive installations.

Both the anti-IED function and its anti-drone counterpart are expected to evolve as new threats emerge. Pending a detailed examination by experts in signal analysis, “the process will consist of recovering the ‘threatening’ component and repatriating it to mainland France and then, in a week and a half, establishing a classifier that will allow to quickly counteract this danger. “.

“It is not the DGA that will weld the components together”, recalls our interlocutor. As the market code prohibits offering this brick to the industrialist, the latter will be supported by DGA MI in the continuous development of its product. It is up to the company in question to integrate the recommendations, specifications and performance levels resulting from DELICE feedback. “This will allow him to take ownership of the technology and for us to obtain a performance that we know is achievable.”

Various avenues of integration are being explored. The duo of the mobile phone acquisition kit is not very robust and does not work well. Therefore, the second phase of the DELICE project will consist of porting the models to other platforms in order to deliver a reinforced and more powerful first prototype. An objective that will depend on several technical and economic parameters, but that could be successful “next year”.

PHP Script, Elementor Pro Weadown, WordPress Theme, Fs Poster Plugin Nulled, Newspaper – News & WooCommerce WordPress Theme, Wordfence Premium Nulled, Dokan Pro Nulled, Plugins, Elementor Pro Weadown, Astra Pro Nulled, Premium Addons for Elementor, Yoast Nulled, Flatsome Nulled, Woocommerce Custom Product Ad, Wpml Nulled,Woodmart Theme Nulled, PW WooCommerce Gift Cards Pro Nulled, Avada 7.4 Nulled, Newspaper 11.2, Jannah Nulled, Jnews 8.1.0 Nulled, WP Reset Pro, Woodmart Theme Nulled, Business Consulting Nulled, Rank Math Seo Pro Weadown, Slider Revolution Nulled, Consulting 6.1.4 Nulled, WeaPlay, Nulledfire

Back to top button