Advanced technologies are increasingly used to commit sophisticated criminal activities. Identifying, preventing and fighting modern crimes demands the implementation of ground-breaking technologies and methods. The EU-funded AIDA project is focussing on cybercrime and terrorism by approaching specific issues and challenges related to law enforcement agencies (LEAs) investigation and intelligence using pioneering machine learning and artificial intelligence methods.
The proposed solution aims to deliver a descriptive and predictive data analytics platform and related tools to prevent, detect, analyse, and combat criminal activities.
While cybercrime and terrorism pose distinct problems and may rely on different input datasets, the analysis of this data can benefit from the application of the same fundamental technology base framework, endowed with Artificial Intelligence and Deep Learning techniques applied to big data analytics, and extended and tailored with crime- and task- specific additional analytic capabilities and tools.
The resulting TRL-7 integrated, modular and flexible AIDA framework will include LE-specific effective, efficient and automated data mining and analytics services to deal with intelligence and investigation workflows, extensive content acquisition, information extraction and fusion, knowledge management and enrichment through novel applications of Big Data processing, machine learning, artificial intelligence, predictive and visual analytics. The final solution will reach Technology Readiness Level 7: system prototype demonstration in operational environment.
AIDA system and tools will be made available to LEAs through a secure sandbox environment that aims to raise the technological readiness level of the solutions through their application in operational environment with real data.
AIDA’s Objectives
The objective of AIDA is to significantly enhance LEAs’ capability to combat cybercrime and terrorist activities through innovations in knowledge mining, information fusion, Artificial Intelligence techniques and analytics services. Envisioned services include:
Detecting and monitoring
relevant data sources for CC and CT investigations in Surface, Deep Web and Darknet;
Mining
huge amounts of heterogeneous data and data sets (text, images, videos, geospatial intelligence, communication data, traffic data, financial transactions related data, etc.) from open sources (e.g., social media platforms, blogs, forums, etc.) and privileged sources (e.g., information collected during a national investigation on a target);
Transforming the data
(structured or unstructured) into actionable intelligence;
Detecting patterns
linked to abnormal/anomaly behaviour, suspicious events, criminal intentions of individuals and groups that would not be visible without AI capabilities;
Predicting
emerging cybercrime and terrorist trends and activities, informing policy-making as well as law enforcement;
Extracting and summarizing
multi-language conversations;
Revealing
hate speech, denigration and disinformation contents;
Linking
investigative relevant contents with the original authors;
Discovering
criminal and terrorist online communities and groups.
Ensuring
a standardised chain of custody for further forensic analysis; and
Enabling
cross-case correlations.
Use cases
Cybercrime- use case
Aim:
To address cybercrime prevention and to speed up cybercrime identification, in order to prepare and deploy the appropriate countermeasures against perpetrators based on crime indicators and trends and leveraging on efficient and effective automated services and novel applications.
Crime and criminal activities:
A broad range of high-tech crimes such as: Malware (code creation and distribution), Ransomware, Hacking, Phishing, Intrusion, Identity theft and Internet related fraud, as well as multidisciplinary cyber-assisted crimes.
Expected benefits for LEAs:
The volume and diversity of the data acquired during seizing infrastructure or wiretapping network communications can take a multitude of months in order to be processed and analysed even at a first layer approach. In the meantime, volume and velocity of open sources’ transmission sets an enormous barrier to timely analyse and correlate data to available information.
Benefits for LEAs will be reached by:
- Applying a direct layer of analysis at ingestion phase, near-real time peeking in the data source will be achieved;
- Automating the data pre-processing parts of the data pipeline we can optimise the dominant stage in time consumption;
- Implementing a framework that incorporates collaborative AI techniques and methodologies a LEA can contribute and benefit at the same moment from every other LEA domain expert.
Counterterrorism - use case
Aim:
Terrorist individuals and groups rapidly adapt their online modi operandi to digital transformation, and rapidly divert their efforts accordingly. LEAs have therefore the hard task to keep up with the rapid changes, predict potential terrorist uses of the new digital trends and functionalities, and are overflowed with huge amounts of data, which can be either structured or unstructured and come in different formats and languages.
Crime and criminal activities:
The crime area of focus for this use case is terrorism, as defined by the Directive (EU) 2017/541 on combating terrorism; this lists the acts to be defined as terrorist offences under national law, when committed with the requisite intention. These include, for instance, attacks, kidnapping or hostage-taking, directing a terrorist group or participating in activities to knowingly contribute to its criminal activities, public provocation to commit a terrorist offence, recruitment for terrorism, providing and receiving training for terrorism, etc.
Expected benefits for LEAs:
There are 3 key expected benefits for LEAs related to the use of AIDA in a counter-terrorism case.
- Increased efficiency - the shorter timeframe required from the start of the investigation to the arrest of the suspect(s), automatic analyses, cross-checks and reports will require a critically shorter time than manual work.
- Increased accuracy - Although AI cannot replace completely humans, particularly when it comes to decision making for security purposes, it will undoubtedly limit human errors in repetitive and simple tasks.
- Stricter ethical considerations - Cultural bias and prejudice can be either conscious or unconscious, and both are human traits that may conduce to unfair treatment and judgement.