Unveiling the TENSOR Project: Transforming Law Enforcement with Biometric Solutions

Authors: IDMG-FUJ-CERTH-SQD

Introduction

TENSOR aims to strengthen public safety in the EU by providing Law Enforcement Agencies (LEAs) with a trustworthy, automated, robust, privacy-preserving, and scalable platform that leverages modern biometric technologies to enable the biometric evidence extraction, fusion, storage, and sharing. Biometric technologies rely on the distinctiveness of a person’s physiological or behavioural characteristics to offer accuracy in the identification process.  Consequently, LEAs have widely adopted biometric technologies for investigations which, however, involve increased complexity in matching the evidence due to potential distortion, loss or enormous volume of data. In order to tackle such complications and to provide security practitioners with improved tools, TENSOR project aims to the adoption of emerging biometric technologies that support forensic investigations, the exploitation of privacy-preserving and legally compliant technologies that fully utilize the potential of behavioural and physiological biometrics for the derivation of concrete and robust contextual evidence, and the formation of a secure, cross-border biometric data exchange system with the use of an accountable and sovereign blockchain mechanism that also ensures interoperability among different legacy systems. Additionally, TENSOR targets to validate the effectiveness and accuracy of the system with real use cases as well as with the LEAs engagement and to disseminate the solutions to forensic institutions for rapid adoption.

Project Overview

The TENSOR project aims to develop and extensively evaluate holistic solutions for security practitioners that enable the easy, secure, and seamless storage and sharing of biometric information and derived intelligence in a harmonized, interoperable, and standardized manner. The project includes the automation of biometric information extraction from various data, identification, identity verification, intelligence, investigation, and evaluation processes, leveraging Blockchain technology for secure, traceable cross-border exchange of biometric information across Europe.

The key partners involved in the TENSOR project include Research Institutions and R&D companies, Police Authorities, other relevant authorities, a wide range of stakeholders, LEAs, and Forensic Institutes across the participating countries. These partners play a crucial role in the piloting and evaluation phases, contributing to the development, testing, and validation processes of the TENSOR solutions.

The main areas of focus within the TENSOR project include modern biometric solutions tailored for security and forensic applications, real-life piloting and evaluation of the solutions through LEAs, innovative technology development, training of practitioners to achieve capabilities upscaling, and biometric data protection during cross-border exchange. These focus areas are designed to advance the field of biometrics and enhance the operational capabilities of LEAs and Forensic Institutes, demonstrating TENSOR’s commitment to innovation and security within the realm of biometric technology.

Biometric Solutions in Law Enforcement

Solving crimes means identification of suspects. Usually, suspects leave numerous traces at a crime scene, including fingerprints, blood traces, DNA, and more. With the emergence of digital technologies, traces stemming from CCTV cameras, smartphones, social media, telecommunication data, confiscated computer devices and storage media are getting more important. In the scope of TENSOR, fingerprints, faces and voices form key objects of interest as these biometrics are unique and thus help to identify suspects. Respective biometric recognition technologies that compare biometric data belong today to the standard repertoire of LEAs.

The analysis of digital data, although very useful, creates challenges as the volume of information to be analysed for each crime increases exponentially. Taking the example of terrorist attacks, several thousands of hours of video material from CCTV cameras plus data from other sources have to be analysed and connected on a short time scale. Furthermore, the matching process requires the facial image to be compared against databases with millions of entries. This results in enormous time pressure, exceeding the capacity of manual management with the current human resources.

It is evident that technical tools are necessary to support LEAs during the investigative work. The aforementioned biometric technologies like face, fingerprint and voice recognition technologies are a key in situations where large volumes of data must be analysed. This technological spectrum will be extended in TENSOR by new approaches for person re-identification, descriptive searches, behavioural biometrics and more.  A key advantage of biometric technologies is their ability to significantly speed up processing and increase efficiency. Furthermore, investigators’ attention wanes over time. Relevant events in a crime scene may also occur in the background of a video, potentially escaping the investigator’s attention. In summary, biometric technologies contribute to a more thorough and accurate analysis.

Key Components of the TENSOR Project

In total, there are 12 core technologies that are developed and utilized within the TENSOR project including physiological biometrics, i.e. face, voice and fingerprint recognition, behavioural biometrics, i.e. gait and patterns recognition, explainable AI, Biometrics Dataspace, Smart Contracts for data sharing, homomorphic encryption, Large-scale Indexing, DSS and the unlocking of mobile devices.

The combinations of the above mentioned modalities lead to the formation of four key offerings of TENSOR:

(1) the first Biometric Dataspace safeguarded by Privacy Enhancing Technologies for a reliable and sovereign data exchange ecosystem,

(2) multiple criteria decision making by ranking of the collected evidence,

(3) innovative application of LLM and generative AI for insightful case story narratives, and

(4) unlocking mobile devices for utilization in forensic investigations.

The impact assessment methodology regarding privacy, ethical, social, and legal concerns in TENSOR is established by four factors: Artificial Intelligence (AI) Impact Assessment, Privacy and Legal Assessment, Ethical Assessment and Social Assessment. Regarding AI, the European Commission established the AI High Level Expert Group (HLEG) which introduced the Assessment List for Trustworthy AI (ALTAI) Methodology and Ethics Guidelines by outlining 7 key requirements that AI systems must meet to be considered trustworthy. Privacy and Legal Assessment focuses on three factors including the identification of the data lifecycle, the analysis of threats prevention, and the specification of the users’ acceptance and adoption of the system. In terms of Ethical Assessment, TENSOR activities adhere to Responsible Research (RR) principles, promote the co-design of the system by engaging the perceptions of users, and strictly comply with EU legislation. Finally, Social Assessment methodology is evaluated for the concepts of identification and authentication, biometric adoption, biometrics acceptability and acceptance, and the acceptance of the TENSOR solutions.

Impact and Benefits

The TENSOR project is poised to significantly transform law enforcement practices through its contributions covering SOTA and innovative biometric technologies integration. By facilitating the digitization and standardization of investigation processes, this initiative promises to enhance the efficiency and accuracy of criminal investigations. By integrating advanced biometric technologies and biometric-based capabilities, this project aims to improve the ability of LEAs to quickly and accurately identify criminal suspects, using digital traces. Such technological advancements are expected to streamline investigative procedures, reduce the time required to solve cases, and increase the likelihood of accurately pinpointing suspects based on digital evidence. Furthermore, the TENSOR project’s initiative to create a biometric data sharing platform accessible by different organizations and spanning across countries represents a significant step forward in the fight against international crime, contributing to public safety and security.

Potential benefits of the TENSOR project for LEAs, stakeholders, and society at large are multifaceted and profound. LEAs are expected to experience an increase in operational efficiency and accuracy in criminal investigations. For stakeholders involved in the security and justice sectors, the project promises improved collaboration and information sharing, both domestically and internationally, fostering a more unified approach to combating crime. Society as a whole stands to gain from increased safety and security, as the project’s initiatives aim to effectively tackle both local and international crimes.

Challenges and Future Outlook

The TENSOR project, aimed at utilizing biometric technologies for solving crimes, confronts several challenges. Firstly, it must adhere to EU’s strict data protection laws like GDPR, which demand careful handling of sensitive biometric data, emphasizing the need for consent, data minimization, and processing limitations. Also, the accuracy of biometric systems is crucial but can be affected by environmental conditions, low-quality data, and physical changes in individuals, necessitating the development of reliable algorithms. Ensuring interoperability and standardization across various biometric systems is also essential for efficient information sharing, requiring global standards for data formats and protocols. Another topic that we should address is potential biases in algorithms, in order to prevent unfair treatment of certain demographic groups is critical. Similarly, the security of biometric data is paramount to prevent unauthorized access and leaks, highlighting the importance of strong encryption and data storage practices. Gaining public acceptance and navigating ethical concerns related to privacy, surveillance, and potential state overreach represent significant challenges.

As outlined, there are several challenges that are part of TENSOR’s development. To overcome such challenges, TENSOR aims to implement stringent GDPR compliant data governance protocols, by limiting and ensuring the processing of biometric data to specific, lawful purposes. Regarding accuracy of biometric data and models, which play a major role in TENSOR’s use cases, continuous update and refinement of the biometric modules are planned as well as the enhancement of the AI solutions, to increase their reliability. Interoperability and standardization will be ensured by setting standards for biometric data formats and communication protocols.  Addressing bias and ensuring fairness, biometric models with diverse datasets will be developed, which accurately represent different demographics to minimize algorithmic bias. Finally, TENSOR aims at transparent communication about the use, benefits, and safeguards of biometric technologies in crime solving and at establishing clear guidelines and oversight mechanisms to prevent misuse and ensure that the use of biometric technologies is always justified.

The outlook for biometric solutions in law enforcement is promising and expected to evolve significantly in the coming years. Several biometric technologies, among other modalities, are increasingly being adopted by LEAs around the world to enhance their ability to identify individuals quickly and accurately. Such developments are likely to shape the future of biometric solutions in law enforcement. First of all, as facial recognition technology advances, its accuracy and the speed of processing are expected to improve, making it even more useful for real-time identification in public spaces, while developments in 3D and AI-powered analysis are making it harder for individuals to evade detection through disguises or other means. Moreover, integrating biometric solutions with existing surveillance systems of LEAs could help with monitoring public spaces, identifying suspects in real-time, and even finding missing persons more efficiently. Finally, the use of mobile biometric devices by police officers in the field is expected to grow, since they can help them to quickly identify individuals removing the need for processing at a station, thereby improving efficiency and reducing the time required to confirm identities.

Conclusion

The core of TENSOR project is described in this article by outlining the main concepts, technologies, and solutions that aim to provide holistic solutions to LEAs, to enhance the effectiveness and accuracy of investigation processes. This improvement is bolstered by the adoption of emerging biometric technologies and the expansion of the technological spectrum in this project, to leverage the digitization of biometric data in a crime scene and utilize it for the purpose of criminal suspect identification.  The main technological contributions of the project include the Biometrics Dataspace, Multiple Criteria Decision Making, Generative AI and Unlocking mobile devices. TENSOR is expected to transform the law enforcement processes with its two key components of the development and integration of biometric technologies for data extraction from mobile devices and the formation of the first Biometrics Data Space. The anticipated outcomes include the improvement in terms of speed, accuracy, and efficiency for lawful evidence data collection and, therefore, successful identification of criminal suspects. Moreover, the project aims to achieve trustworthy, automated, privacy-preserving and sovereign information sharing, which is expected to enhance efficient collaboration and cooperation in cross-organizational and cross-border investigation scenarios. Both technology offerings rely on SOTA biometric technologies and adhere to Responsible Research principles.