Recordings of our voice are increasingly being captured, stored and processed when we interact with voice driven interfaces and other automated services. For the most part, this is done without any nefarious intent. Even so, recordings of our voice contain inherently sensitive, personal information, information that should not be willingly entrusted to others. Examples include health and socio-economic status, geographical background, ethnicity, personality and emotion, in addition to information concerning our social circles, family and relatives. Since this information can be exploited for ethically reprehensible purposes, safeguards are required to prevent privacy infringements. There are two general strategies: data protection or anonymization. While the best solution is naturally application dependent, anonymization can be more flexible and cost efficient. Anonymization techniques can be used to strip speech signals of their personally identifiable information while retaining intelligibility and quality. Anonymized speech signals can then be processed and stored without the possibility of information gleaned from them being matched to the speaker. Unfortunately, there are few solutions and progress in the field is hampered by the lack of open datasets and open tools. These are critical for assessment and the meaningful comparison of competing approaches. As a problem that requires pattern recognition techniques for assessment, the lack of such datasets and tools is a significant barrier. Harpocrates will form a working group / implementation network that will not only collect and share the very first such open datasets and open tools, but will also launch, as part of the project, the first open challenges in anonymization. Open challenges will fuel progress and ensure that emerging technologies are transferred rapidly to industry which must meet increasingly stringent demands for privacy preservation. Anonymization will then form a critical component in delivering regulatory compliance and a critical consideration in privacy-by-design methodologies.

Kinetic models are accurate descriptions of interacting particle systems in physics. However, their numerical resolution is often too demanding, as they are defined in the large-dimensional position-velocity phase space and involve multi-scale dynamics. For this reason, reduced models have been developed that represent optimal trade-offs between numerical cost and modelling completeness. In general, this reduction is carried out in two ways. The first is based on asymptotic models that filter out fast dynamics and are obtained when a small parameter tends towards zero (collision/oscillation limit). The second, called reduced order modelling, consists in finding a smaller representation of the problem able to describe the dynamics (POD). The main objective of this project is to design new reduced order models that are more efficient than classical ones, based on machine learning techniques applied to kinetic data. Ensuring the stability of the models obtained will be a key point studied

Ischemic stroke is a major cause of disability and death worldwide. The induced severe disability cannot be reversed but with rapid treatment action. Recent clinical trials have demonstrated Endovascular Thrombectomy (EVT) to be highly effective, which has led to its widespread adoption in clinical routine. However, the conditions for a safe and efficient treatment are a strict management timeline and highly expert interventionalists. Training is a key element of performance and process improvement, and simulation plays an increasing and essential role, but there still lacks evidence that the commercial simulators can help improve clinical performance. In other words, the experience they convey is hardly transferable to the intervention room. PreSPIN (Predictive Simulation for the Planning of Interventional Neuroradiology procedures) aims at bridging the gap between training and intervention by addressing the planning phase. Here, numerical simulation can play a fundamental role on the condition that it is predictive, i.e. capable of rendering events that are critical to medical doctors. The project brings together a multidisciplinary team of researchers in numerical simulation, image processing and clinical medicine in order to tackle theoretical, technological and medical issues standing in the way towards a predictive planning system for the therapeutic management of acute ischemic stroke. The overall objective is to develop original methods to allow for high-fidelity and interactive simulation of catheter navigation and placement in the intracranial circulation, and real time synthesis of perfusion MRI images that takes advantage of an accurate simulation of blood flow in both the large vessels, as well as the capillary brain tissue. New blood vessel models will be designed, dedicated to the considered numerical simulations, and able to accurately capture the complex topology and geometry of the tortuous brain vasculature. In that context, PreSPIN will: • Objective 1: develop new solutions for segmenting and modeling the brain vasculature of the patient from MRA data, to provide geometric boundary conditions to forthcoming simulations; • Objective 2: propose new blood vessel models adapted to interactive simulation of interventional devices navigation, for a more informed choice of medical devices and treatment options; and fast blood flow simulation, to prevent potential risks associated with blood flow restoration; • Objective 3: progress in computational fluid dynamics (CFD) and investigate new and refined means of simulating perfusion imaging, to better predict treatment outcome. These advances will be supported by experimental data acquisitions with the aim of providing open-source in vitro data and reproducible setups to the community. Such data will be designed in close collaboration with interventional neuroradiologists who belong to the project’s partner teams, to both ensure critical phenomena are captured and make advances towards an engineered definition and assessment of medical predictivity. The project will have a definite impact on the clinical use of simulation for planning, but also for post-operative case review, which will constitute a situation of choice for our validation purpose. The results will directly be applicable to training and will enable a seamless integration of patient data in simulators, which is a major achievement to bring training experience closer to actual clinical conditions and make it transferable to the intervention. A longer term objective is to leverage simulation during the intervention to complement intra-operative data with unseen simulated information (e.g. blood pressure, device friction force, or full 3D view).

Formal deductive methods aim at improving the quality of software by relying on tools based on strong mathematical grounds, like set theory. Those tools allow the user to prove the properties that ensure the correctness of a software with respect to its specification. When human safety is at stake, confidence in proofs is crucial, however the implementation of a prover is a complex, error-prone, and sometimes buggy task. ICSPA focuses on the set-based specification formalisms B, Event-B, and TLA+ that have been adopted for industrial development projects, for applications where correctness is critical. It aims at reinforcing the confidence in proofs carried out mechanically using them. Our project also aims at designing and implementing an exchange framework, through which those three systems can share their proofs and theories, making them effectively interoperable. The strategy we adopt is to verify these proofs formally and independently with Dedukti, a proof checker simple enough to be audited manually or even re-implemented. Through the versatility offered by Dedukti as a common foundation, safe-software programmers will be able to exchange artefacts between B, Event-B and TLA+ developments, within their respective IDEs Atelier B, Rodin and TLAPS. This approach is inspired from the one successfully instigated by Logipedia, a small library of mathematical results that can be exported to, and verified by, a wide range of formal systems, including Coq and HOL. To reach our goals, we will express the set theories underlying the three specification languages in Dedukti and then export their proof traces, in order to verify them independently using Dedukti. This will be the core mechanism for a more ambitious export of complete models used for the development of real software for safety-critical systems, with a particular focus on Labelled Transitions Systems. With this data, we will define translations between systems at the Dedukti level and import functionalities into the tools Atelier B, Rodin, and TLAPS. This will allow for importing in one system what has been exported from another system. This backbone architecture will be supported by proof reconstruction features provided by automated theorem provers, that will enable explicit completion of high- and middle-level, often incomplete, proof traces. It will be backed up with additional anchors, obtained by horizontally matching definitions and statements across tools. The impact of this methodology will be assessed on a very large body of proof obligations provided by our industrial partner. Moreover, case studies for interoperability will be elaborated. Lastly, beyond academic and educational/training dissemination, the export/import, completion by an automated theorem prover, and verification by Dedukti features will be integrated into Atelier B and exploited at an industrial scale.

Whilst the technology of Unmanned Aerial Vehicles (UAVs) is highly advanced for outdoor flight, the realm of confined flight poses a significant challenge due to the complex interplay of aerodynamic couplings and interferences between the UAV and surrounding walls. These disturbances can wreak havoc on the UAV's stabilisation controls and hinders also its manœuverability. Our project endeavors to overcome this challenge by combining aerodynamic approaches, biomimetism and machine learning to enhance the control and stability of UAVs in confined and near-wall environments. Our research will focus on both rotor and flapping-wing drones (ornithopters), and we plan to augment the traditional control strategies, which primarily rely on closed-loop response from embedded kinematic data. To achieve this, we aim to better account for (i) the induced aerodynamic perturbations, (ii) the drone's proximity to the walls, and (iii) bio-inspired solutions derived from the study of butterfly flight in confined spaces and near walls. To achieve this ambitious goal, our multi-disciplinary consortium gathers experts in robotics, biorobotics, fluid mechanics, and entomology, alongside an industrial partner (XTim) renowned for its leadership in the market of biomimetic drones with flapping wings. Our project is structured around several scientific and technical tasks that aim to finely characterize the dynamics of confined flight and aerodynamic perturbations for both multi-rotor and flapping-wing UAVs, identify the free and confined flight characteristics of butterflies for biomimetic purposes, and develop innovative control protocols (based on reinforcement learning) to achieve stable UAV flight in confined or near-wall environments. These tasks rely on a combination of dedicated experimental facilities, field observation of butterfly flights, state-of-the-art aerodynamic metrology and simulations, and advanced machine learning algorithms, all developed within the consortium. The culmination of our efforts will be the synthesis and validation of our findings, which we will implement on our industrial partners' commercial UAVs. Our developed solutions and control strategies have the potential to revolutionize the field of confined UAV flight and enable safe, reliable and stable operation in areas that were once considered inaccessible.