NOTE! This site uses cookies and similar technologies.

If you not change browser settings, you agree to it. Learn more

I understand

Learn more about cookies at :

November 5, 2020 | Image recognition system can classify, annotate, and explain how

reconaissance images 250

Explainable AI—when an artificial intelligence produces an explanation of a decision along with the decision itself—is a requirement for certain applications. CEA List, a CEA Tech institute, recently tackled this major issue in trusted AI by developing a new machine learning module that can classify images and annotate objects. The module has been integrated into the ExpressIF® AI platform.

Increasingly, artificial intelligence is used to make decisions that affect our day-to-day lives. In this context, trust is vital. And to trust an AI's decision, you need to know how the AI arrived at that decision. In PhD research co-supervised by the CentraleSupélec MICS Lab (which studies mathematics and computing for complexity and systems), the Carnot CEA List developed a new machine learning module that can classify images and annotate objects and generate an explanation at the same time. This module has been integrated into the CEA List symbolic AI platform ExpressIF®.

Here's how it works. First, a neural network "understands" an image by identifying specific areas that correspond to different objects in the image. The new algorithms integrated into ExpressIF® then take over, identifying the objects according to their relative positions, and then annotating them. What makes the algorithms so powerful is that they can learn to identify objects error-free from just a few (fewer than ten) images. The initial tests—carried out on abdominal MRI images—were encouraging. Not only was ExpressIF® able to automatically annotate the organs pictured, but it was also able to justify its annotations with an explanation produced in natural language. This is a clear step forward toward helping doctors interpret their patients' scans.

The solution augments users' trust in the AI's decision, of course. But it also aligns with future laws, which will likely require explainable AI for certain applications. Here, the researchers tested the new feature on medical images. However, it could also augment scene interpretation or the characterization of manufactured parts, for example.

Read article at




November 3, 2020 | CEA and Systerel: a new step towards the functional modeling of the interlocking system of the future

CEA002405 HDThe CEA List Institute and Systerel have reached a major milestone in the digitization of railway systems, by delivering a new version of the functional modelling module for new-generation of interlocking systems to SNCF Réseau. They are supporting the growing complexity of these systems with an efficient modelling and validation environment.

In 2018, SNCF Réseau issued a request for proposal for the development of the "Atelier Métier Signalisation" (AMS), dedicated to the specification and formal validation of Computer-Based Interlocking (CBI) systems. The CEA List institute and SYSTEREL company response to the request was reviewed and accepted by SNCF Réseau in June 2019 as the most relevant one regarding the technical proposal as well as the expertise of both companies: modelling and validation of complex systems (CEA List) and proof of safety properties on critical systems and railway domain (Systerel).

To simulate operating scenarios in complete safety

AMS development is divided into three main phases, which will ultimately materialize in the production of a tool enabling the specification (i.e., the formal description) and the validation (through simulation and formal methods) of CBI systems.

During the first phase, CEA List and SYSTEREL have designed and implemented the AMS modeling capabilities. These developments are grounded to Papyrus1 a modeling tool developed by CEA List for 10 years. The proposed environment integrates all the functionalities required to model and instantiate CBI systems. In particular, it enables SNCF Réseau to:

  • Specify the CBI system behavior and parameters,
  • Import the parameters provided by SNCF partners involved in the development of CBI systems.
  • Automatically instantiate a CBI system according to a set of parameters and a behavioral specification.

During the next months, Systerel will add features enabling formal verification of CBI systems as well as the track plans’ edition. The combination of these features with the simulator developed by CEA will especially provide the capability for railway signaling engineers to understand the conditions in which safety properties do not hold and then proceed to the appropriate refinements in the system specification.

An open-source tool for optimal adaptability

The AMS source code will be released as open source at the end of each phase, i.e. by the beginning of 2021 for the AMS modeling capabilities. The entire source code will be accessible by 2023.

As part of the ARGOS2 innovation partnership, the AMS will enable SNCF Réseau to reduce time and costs required to develop and maintain new generation CBI systems. The AMS will contribute to increase the interoperability of these large-scale systems, capable of controlling all signaling equipment with the agility needed to adapt to the track plan, whatever its complexity.
2The Argos innovation partnership launched by SNCF Réseau in 2018 aims to develop the 3rd generation of interlocking systems, in collaboration with manufacturers Alstom, Hitachi-Ansaldo, Siemens and Thales.

About CEA List

The CEA List, located in the CEA Paris-Saclay and CEA Grenoble centers, is a CEA technological research institute that helps its industrial partners to enhance their competitiveness through innovation and technology transfer. Focused on smart digital systems, its R&D programs focus on artificial intelligence, advanced manufacturing, cyber-physical systems and digital health. The CEA List is a member of the Carnot Institutes network.

More info: | @CEA_List | LinkedIn | YouTube

About Systerel

Systerel is an independent company, created in 2002, whose core business is the development, validation or evaluation of critical real-time or safety systems. Its recognized technological expertise, mastery of SIL processes, knowledge of railway signaling and its ability to develop innovative tooling solutions have led to the existence of privileged relationships with major players in the railway market such as Alstom, Siemens, Hitachi-Ansaldo, RATP and SNCF. | LinkedIn


October 30, 2020 | Protecting cryptographic code against new cyberattacks

protection code cryptographiques 250

CEA List, a CEA Tech institute, developed a module to prevent a new—and increasingly common—cyberattack known as a temporal attack. The module was presented at the IEEE Symposium on Security & Privacy 2020, and has been integrated into the BINSEC code analysis suite. During testing on real programs, it proved effective against these attacks.

For years, the Carnot CEA List, has been using tools that leverage formal methods to verify that there are no bugs in a program. These tools are very effective against conventional cyberattacks. However, they don't work as well on side-channel attacks. These attacks take advantage of the physical phenomena that occur when a program is run, rather than vulnerabilities in the program itself, to extract information. Temporal attacks are designed to decode cryptographic keys by comparing the processing times for different operations on a computer, for example.

A technique called "constant time" coding can be used to fend off temporal attacks. The idea is to make sure that the execution time for a given task—such as the verification of a password—is not dependent upon the input data. Because this type of coding is so tricky, CEA List researchers developed REL, a tool that can automatically verify that code is "constant time." The tool has been integrated into the BINSEC code analysis suite and came through tests on 338 cryptographic implementations with flying colors.

The advance, which was presented at the IEEE Symposium on Security & Privacy 2020, will pave the way toward new code analysis methods for other kinds of attacks and, more generally, better and more automated security for cryptographic programs.

Read article at



9 octobre 2020 | Lesly-Ann Daniel, doctorante au CEA List reçoit le Prix 2020 Jeunes Talents L'Oréal-Unesco pour les Femmes et la Science


A l'occasion de la fête de la science 2020, Lesly-Ann Daniel, 25 ans et originaire d'Orléans, a reçu le prix Jeunes Talents décerné par L'Oréal et l'UNESCO pour ses recherches en informatique.

Lesly-Ann Daniel découvre tardivement l’informatique et s’enthousiasme pour le codage. Concevoir des programmes pour résoudre des problèmes, les faire tourner sur son ordinateur et pouvoir tester immédiatement sa solution l’enchante : la plasticité de l’informatique lui apparaît comme un atout extraordinaire.

Alors qu’elle est étudiante à l’université de Limoges, elle s’intéresse à la sécurité informatique, qui devient par la suite son principal domaine de recherche. Aujourd’hui en deuxième année de thèse, Lesly-Ann Daniel conçoit des logiciels capables d’analyser automatiquement des programmes pour trouver des failles de sécurité ou pour garantir leur absence. Elle analyse en particulier les logiciels cryptographiques afin de s’assurer qu’un potentiel « pirate » ne puisse en extraire des informations confidentielles, ce en mesurant le temps d’exécution de ces programmes.

Pour Lesly-Ann Daniel, le Prix Jeunes Talents s’inscrit dans la continuité d’un parcours marqué par la ténacité, en dehors des filières d’excellence des classes préparatoires et des grandes écoles.

C’est pour elle « une formidable opportunité d’apprendre à croire en soi et en ses capacités, d’autant plus nécessaire lorsqu’on est une jeune femme dans un domaine comme l’informatique où le féminin demeure minoritaire ».

La chercheuse, passionnée de lecture et de cuisine, s’adonne également à l’escalade et à la randonnée. Elle apprécie particulièrement d’assister à des conférences ou de se rendre dans d’autres laboratoires que le sien pour se nourrir d’interactions fertiles entre chercheurs.


Pour aller plus loin :

October 5, 2020 | Advanced ictometer for nuclear reactor maintenance

montage visuel 250
© Adobestock

​In research conducted for a nuclear-industry equipment manufacturer, CEA List, a CEA Tech institute, developed a new ionizing radiation measurement device. Built on the most advanced technology around, the device was designed to meet the requirements of use in nuclear power plants.

Operators who perform maintenance on nuclear reactors use ictometers to measure the neutron flux around the reactor. Strategically placed sensors transform the neutrons detected into electrical pulses, which are in turn centralized and counted by the ictometer. As the older generation of ictometers reaches obsolescence, CEA List, a member of the Carnot Network, drew on its know-how in several fields to design a completely new measurement system for an equipment manufacturer. The powerful, integrated device was developed to meet today's requirements.

CEA List leveraged the know-how available at its PACT sensor and processing lab to develop a multi-channel digital ictometer that can measure radioactivity at several (in this case, four) points simultaneously. CEA List also designed the device's acquisition circuit, which transforms current pulses into voltage pulses, which are then digitized, analyzed, and counted to reliably and reproducibly generate usable data for operators.

The new ictometer also has some additional added-value features like data logs covering several days, charts that make tracking neutron fluxes easier, and a specially designed human-machine interface that displays results in a way that is useful to operators in the field. Other features, such as the integration of step-by-step operator instructions are already on the drawing board.

This end-to-end system does everything from measure to generate actionable data. Not only is it a fine illustration of CEA List's comprehensive digital instrumentation know-how, but it also provides nuclear-industry equipment manufacturers with an innovation that will change the way measurements are taken in the field.

Read article at