Recently, Spikenet has passed a test in real conditions with cameras of CCTV in Paris. The faces of 150,000 individuals in reason 5 to 200 faces by individual (1.5 million images in total) have been loaded in the software in order to find one person. Like looking for a needle in a haystack. "In a few seconds, the face of the person sought was released 8 times on the first 20 results", resumes Hung Do - Duy bringing its algorithms on hardware platforms for three years.
Article here and translated below. posted on the 20th of April:
https://business.lesechos.fr/entrep...tart-ups-de-l-ia-et-de-la-securite-308786.php
Artificial intelligence (deep learning / machine learning) allows to identify and stop a cyber attack, then to repair its damage. It is also used for the recognition of faces in a crowd. Numerous start-ups emerge in this innovative sector.
Security and cyber security are areas strong consumers of artificial intelligence. Evidenced by Emmanuel Meriot, Director in France of the British Dark track, which the Antigena solution has equipped, for four years, 2,000 companies worldwide. "Our algorithms provide an immune system against cyber attacks. Especially, since last year, they quit the current attacks. "Including those of Ransomware (Ransomware) that encrypt data drives infected before the ransom demand. These algorithms are embedded in machines that analyze the normal data traffic in the corporate network in order to create a reference model. As soon as an anomaly, even in the form of weak signals, it will be detected and will generate an alert. Above all, for example in the case of a Ransomware, suspicious encryption of hard disks will be stopped from outset. After the crisis management, the administrator may investigate (forensics) for the trace of the criminal.
To read also
The French Tech seduced by cyber security
Auxiliary also the fight against terrorism, solutions 'machine learning' want to talk video images. "Our software understands and analyzes what is happening in an image within a real-time stream," explains Stéphane Bidault, president of TEB (17 million euros of turnover in 2016, 180 employees). Then, the system extracts useful information: counting people passing in front of the camera, analysis of objects (package abandoned, for example), automatic reading of license plates, detection of abnormal crowds or people. The images are then transformed into data that will be interpreted from a normal situation to trigger actions. As the pre-positioning of contingency forces or guiding the flow of people to evacuate.
On their side, the Toulouse Spikenet neural networks are used, among other things, by Homeland Security (SID) branch and the prefecture of police of Paris. The algorithms manage a tour de force. "Their learning doesn't have to be constantly supervised by humans. Our neural networks to recâblent all alone, ensures Hung Do - Duy, president of Spikenet Technology SAS who works for eleven years with the Ministry of the Interior. Our research on unsupervised neural networks began twenty years ago. »
A.I.Mergence launches an autonomous security robot
Recently, Spikenet has passed a test in real conditions with cameras of CCTV in Paris. The faces of 150,000 individuals in reason 5 to 200 faces by individual (1.5 million images in total) have been loaded in the software in order to find one person. Like looking for a needle in a haystack. "In a few seconds, the face of the person sought was released 8 times on the first 20 results", resumes Hung Do - Duy bringing its algorithms on hardware platforms for three years.
To read also
Cybersecurity: Datadome get 1 million euros and aims Europe
"We started on a Raspberry Pi [of nano-computers the size of a card of credit less than 50 euros, Editor's note]. Now, we realize chips electronic neuromorphiques on reprogrammable integrated circuits (FPGA). "A strategy identified by the Brainchips American who bought the Toulouse company in September. What build networks of neural networks by integrating these chips for servers, networking equipment, cameras, robots...
Finally, the French start-up
A.I.Mergence is about to launch an autonomous mobile robot for the safety of the buildings. Thanks to its algorithms 'machine learning' and 'deep learning', the robot adapts to its environment and recognizes the face of the occupants and distinguishes them from unauthorized people. In case of intrusion, it sends a 95 dB siren and blinding flashes. This robot will be available in presale in June next on a crowdfunding site.
(20min delay)