Sécurité numérique

Deepfake detection is technically mature but legally unusable: no current system explains why audio is classified as synthetic. This project delivers forensic-grade explainability and reproducible benchmarks, unlocking adoption in forensics, law enforcement, and finance.

Challenges related to “on the move” biometrics are 1/ lower quality live biometric data, and 2/ no time to read the ePassport. Also, fully automatic biometric border control solutions, even with a stop allowed, are currently deployed only for pedestrians in controlled environment. Carmen offers biometric solutions for non-stop border control, suitable for pedestrians and vehicles, in uncontrolled environmental conditions. Travellers’ authentication is achieved in two steps: 1) the biometric data (face, iris, periocular) of travellers is securely stored in their smartphones, thanks to a DTC (Digital Traveler Credential).2) the biometric data is securely transferred from the DTC to the police infrastructure, and compared to the live biometric data collected as the traveller crosses the border control point.
To address a variety of environmental conditions, Carmen uses both NIR and RGB live biometric images and compare them to the reference images that have been acquired with one type of lighting only. To make biometrics more robust, Carmen has a multimodal approach on face iris and periocular regions. In the fraud detection, Carmen detects presentation attacks on moving travellers and searches strange travellers’ behaviour. To address small and large border control points, Carmen enables the use of fix and body-worn cameras. To address travellers in cars or lorries, Carmen detects face images in slowly moving vehicles, through the windows. Travellers in coaches are controlled by border guards walking through the coach, using dedicated portable equipment as they walk. Carmen addresses the robustness of DTC via data injection attacks. Of course, Carmen complies with the existing legal and ethical standards, and privacy is a central concern. Carmen solutions will be demonstrated in operational conditions over the UK - France border, as French and UK border authorities are part of Carmen consortium, using the infrastructure proposed by the partner Brittany Ferries.

Along the whole value chain in using data for economic purposes, guidelines and tools are required to make the business of the different stakeholders successful, and the end-users confident that none of their rights are endangered. CERTAIN addresses these needs and delivers solutions for data holders, dataspaces and AI systems providers, and AI systems deployers, which are the primary actors of the data and AI value chain. They must be compliant with applicable European regulations, must reach this compliance in a timely manner, and at reasonable cost.
CERTAIN delivers guidelines and technical tools to help with compliance, to assess data quality, to measure biases in datasets, and to protect privacy. CERTAIN sets the foundation of AI certification: it translates the regulations to business terms, builds a directory of certification entities per business, develops a platform to streamline the certification process, and tools for AI system providers and certification entities so that they could respectively prepare and run a certification process.
In case of security breach, not only privacy may get compromised, but also AI models may become useless and lead to extremely damageable decisions. To make sure that AI-based products are of high quality and reliability, CERTAIN develops security tools and methods, specifically suitable for dataspaces and AI systems.
CERTAIN addresses the environmental footprint of the AI value chain. Innovative techniques are elaborated to reduce energy consumption when building and running AI systems. This is beneficial not only for the green deal but to reduce cost for AI stakeholders.
As importantly, CERTAIN considers the end-users perspective, and provides templates and guidelines that may be used by AI systems deployers to reassure end-users on the use of their private data. The project tests its results on seven operational pilots in six different business areas, considering all the actors along the AI value chain.

DEFORM addresses the urgent need for robust, explainable AI forensic tools to counter the growing threat of Deepfakes, AI-generated or manipulated audio, video, and images that challenge current forensic tools. The project will deliver a modular, practitioner-driven AI framework with multi-level explainability for multimodal Deepfake detection and source tracing, validated (TRL5) and

demonstrated (TRL6) in 2 operational forensic environments. DEFORM’s innovative approach leverages emerging Foundation Representation Models and few-shot learning for robust and multi-level explainable AI (xAI) to ensure adaptability to new manipulation techniques and transparency for diverse stakeholders, including investigators, forensic examiners, legal professionals,

and lay audiences.

The consortium unites leading scientific, technical, legal, and practitioner partners from six European countries, including police authorities and forensic institutes, to co-design, validate, and demonstrate solutions aligned with EU legal, ethical, and forensic standards. By twinning expert research partners in scientific, legal and ethical part from partner countries with practitioners from respective countries, DEFORM outputs include validated detection and source tracing modules for image, audio, and video; humancentred xAI explanations; annotated Deepfake datasets; and open-source forensic tools for wide adoption.

DEFORM’s impact pathway ensures improved reliability and usability for modern forensic analysis using new and emerging technology. This will enhance the capacity for law enforcement to fight against crimes in digital area and strengthen public trust.

DEFORM's innovative training curricula will bridge the gap between research and operational practice. Also, stakeholder engagement will foster wide adoption and contribute to a strong and independent EU AI continent.

Artificial Intelligence is a disruptive technology of our times with expected impacts rivalling those of electricity or printing. Resources for innovation are currently dominated by giant tech companies in North America and China. To ensure European independence and leadership, we must invest wisely by bundling, connecting and opening our AI resources. AI4EU will efficiently build a comprehensive European AI-on-demand platform to lower barriers to innovation, to boost technology transfer and catalyse the growth of start-ups and SMEs in all sectors through Open calls and other actions. The platform will act as a broker, developer and one-stop shop providing and showcasing services, expertise, algorithms, software frameworks, development tools, components, modules, data, computing resources, prototyping functions and access to funding. Training will enable different user communities (engineers, civic leaders, etc.) to obtain skills and certifications. The AI4EU Platform will establish a world reference, built upon and interoperable with existing AI and data components (e.g. the Acumos open-source framework, QWT search engine..) and platforms. It will mobilize the whole European AI ecosystem and already unites 80 partners in 21 countries including researchers, innovators and related talents. Eight industry-driven AI pilots will demonstrate the value of the platform as an innovation tool. In order to enhance the platform, research on five key interconnected AI scientific areas will be carried out using platform technologies and results will be implemented. The pilots and research will showcase how AI4EU can stimulate scientific discovery and technological innovation. The AI4EU Ethical Observatory will be established to ensure the respect of human centred AI values and European regulations. Sustainability will be ensured via the creation of the AI4EU Foundation. The results will feed a new and comprehensive Strategic Research Innovation Agenda for Europe.

Standard machine learning systems require massive data and huge processing infrastructures, but the main limitation to their spreading comes from the need of the empirical and rare knowledge of an experienced data scientist able to set and adjust their behavior over time. The ALLIES project will lay the foundation for development of autonomous intelligent systems sustaining their performance across time. Such unsupervised system will be able to auto-update and perform self-evaluation to be aware of the evolution of its own knowledge acquisition. It should adapt to a changing environment by following a given learning scenario that balances the importance of performance on past and present data to avoid unwanted regression. Such systems could not be developed without adapted metrics and protocols enabling their objective and reproducible evaluation. This evaluation should assess the performance on the given task and quantify the effort required to reach it in terms of unsupervised data collected by the system and of interaction with humans in the case of active-learning. The ALLIES project will develop and disseminate those metrics and protocols. They will be available to european actors via an open evaluation platform dedicated to reproducible research. An evaluation campaign and a workshop will be organised to engage the community on this path. By publicly releasing the evaluation protocols and data, by releasing a dedicated evaluation platform and by developing autonomous systems for two tasks: machine translation and speaker diarization, we believe that the ALLIES project will boost the development of intelligent lifelong learning systems in Europe.

Building, testing and adapting bespoke speaker identification models to be used in IDIAP's 'Bob.Spear' speaker recognition software