Periodic Reporting for period 2 - ROXANNE (Real time network, text, and speaker analytics for combating organized crime)
Berichtszeitraum: 2021-03-01 bis 2022-12-31
ROXANNE (Real Time Network, Text, and Speaker Analytics for Combating Organized Crime) was a research and innovation project funded under the EU Horizon2020 Programme that was officially launched in September 2019. The 40 months long project led by the Idiap Research Institute brought together 25 partners, including 11 LEAs, SMEs and representatives of industry and academia from 15 countries. ROXANNE focused on privacy-aware criminal network analysis enhanced by AI technologies supporting multilinguality (specifically EU languages) and mutimodality (speech, text, video, metadata) to provide LEAs with an efficient legal framework and technical tools to track and uncover criminal identities. INTERPOL’s participation helped to safeguard the developed tools’ compliance with applicable legislation in the area, including with the INTERPOL Rules on the Processing of Data. The technological developments provided in the ROXANNE analytical platform enable LEAs to achieve a significant increase in the speed of investigation processes and an improvement in the identification of individuals in the scope of criminal cases where large amounts of lawfully intercepted communications are analysed.
The work performed over the first 18 months of the project include:
- analysis of requirements from LEAs related to the investigation of (cross-border)large criminal cases
- analysis of ethical, societal and data privacy concerns related to the technological developments in ROXANNE, and development of solutions
- analysis of existing data (i.e. fabricated, simulated or real-data from LEAs) to carry our R&D in AI related tasks
- research and development of audio, video, text processing technologies applicable in the project platform interfaced with the network analysis
- integration aspects including analyses of existing platforms
- new developments in construction and analysis of criminal networks, taking into account the prior knowledge and the evidence extracted using AI-related algorithms
- organisation of three field-test events, together with several stakeholder board meetings
- communication and dissemination activities in several directions (e.g. close collaboration with existing FCT projects such as FORMOBILE and LOCARD)
- analysis of requirements from LEAs related to the investigation of (cross-border)large criminal cases
- analysis of ethical, societal and data privacy concerns related to the technological developments in ROXANNE, and development of solutions
- analysis of existing data (i.e. fabricated, simulated or real-data from LEAs) to carry our R&D in AI related tasks
- research and development of audio, video, text processing technologies applicable in the project platform interfaced with the network analysis
- integration aspects including analyses of existing platforms
- new developments in construction and analysis of criminal networks, taking into account the prior knowledge and the evidence extracted using AI-related algorithms
- organisation of three field-test events, together with several stakeholder board meetings
- communication and dissemination activities in several directions (e.g. close collaboration with existing FCT projects such as FORMOBILE and LOCARD)
ROXANNE has already published (or prepared for publication) 51 scientific (peer-reviewed) documents that are directly or indirectly related to the project objectives. More specifically, this includes audio, text processing and (criminal) network analysis related publications. ROXANNE's primarily goal is to substantially speeding up of the casework investigation (by moving tiring and time-consuming tasks to machines), thus limiting the time the criminals have for planning further strategies, and making them more vulnerable. Our goal is therefore to make high-level evaluations of the technologies developed in the project as it would be expected by the police practitioners. Performance analyses are planned to be done on material which is as close as possible to the real casework of LEAs.