Subscribe to our newsletter
Subscribe to our newsletter for updates from all the latest news of the Project!
TENSOR
Platform
Face and Person biometrics
The component for person tracking/re-identification using person related prominent features is currently at TRL 4, while the estimation of facial characteristics such as gender or age (soft biometrics) is at TRL 5 on video. Through TENSOR, the improvement of both technologies in operational environment and optionally the incorporation of additional characteristics will allow them to reach TRL 6/7.
Voice biometrics
The voice recognition is part of PHO offering (TRL9). Existing enhancements such as voice decoders, deep fakes (TRL4) for system robustness, reliability improvements as well as additional contextual and behavioural evidence and intelligence will be derived through the TENSOR project and will validate these technologies in real setting reaching TRL7.
Fingerprint biometrics
Contactless fingerprint capturing by the means of COTS devices was validated in a relevant environment, on a limited data sample (TRL 5). Through the TENSOR project, the intention is to reach TRL 7 by proving the concept using comprehensive and extensive data samples and achieving automated calibration of COTS devices for this purpose.
Gait biometrics
Through the H2020 FINSEC56 project, an initial work regarding 2D projection of body-joints has already been produced. Through TENSOR, the technology will be enhanced with multi-camera analysis, improvement of joints tracking and enhancement to the project to create a 3D model. The improvement and new development will be demonstrated in relevant environment reaching TRL7.
Behavioural biometrics
The user continuous behavioural authentication is currently at TRL 5 stage and it has been already validated in retail/manufacturing domain. Through TENSOR the technology will be enhanced with additional behavioural patterns including gait and fraud detection (transactions) and other actions, reaching TRL7.
Distributed biometrics data sharing platform
The platform is currently at TRL 5 allowing secure and trustworthy data sharing. Through TENSOR, the platform will be enhanced with biometric data specific exchange capabilities as, improved efficiency in terms of transaction costs and interoperability with legacy systems combined with real setting validation reaching TRL7.
Biometric Data Protection and Privacy
The biometric information protection component is currently at TRL 3. Through TENSOR, the initial design and software package for fuzzy extractor and homomorphic encryption will be further developed, validated and demonstrated in the related use cases reaching TRL7.
Large-scale Biometric Data Indexing
The technology is currently at TRL4 through the ISFP Anti-FinTer project. Through TENSOR, investigation of multi-modal biometric deep hashing architectures, combined with efficient domain adaptation/transfer learning techniques and validation in real setting will lead to TRL7.
Automation and Scalability of Identity, Intelligence and Investigation
Within the scope of TENSOR project, a risk-based Decision Support System will be offered, while the AIaaS functionalities will be integrated within it. The combined suite with the validation in real setting will increase the TRL from 4 to 7 by the end of the project.
Unlocking criminals' devices using Biometric Data
In H2020 ASSURED57, it was developed and validated in a laboratory environment. During TENSOR, the module will allow the extraction of the biometric data on various devices and will be demonstrated in relevant operational environment reaching TRL7.
Responsible and Explainable AI for Fair and Unbiased Biometric Technologies
The XAI module has been partially implemented in LAW-GAME project58 for evidence recognition where LEAs are conducting forensic examination. AI methods including Knowledge Graphs will be tested, combined with ML to enhance the decidability process related to the responsibility, validity and trust of AI systems (TRL7). In the case of TENSOR, XAI will be used to explain to the forensic practitioners the reasoning behind mapping of different biometric evidence derived from the crime scene into a potential suspect. In TENSOR, construction of separate Knowledge Graphs will assist in tracing the decision-making process and thus make it easier to the Forensic Practitioners, to understand how each of the overall system reach the proposed outcomes.
European Biometric Data Space
TENSOR will introduce the first common European Biometric Data Space that will improve the exchange and access of biometric information (biometric profiles, more or less distinctive biometric features, derived intelligence) among security practitioners. The ultimate goal will be to improve the European investigation capabilities towards fighting terrorism and other forms of organised crime (primary use of data) and for biometric/forensic research and policy making purposes (secondary use of data).
