COPKIT is developing data-driven policing technologies to support Law Enforcement Agencies in preventing, investigating and mitigating crime and terrorism.
In this blog, the University of Granada provides an overview of COPKIT’s Knowledge Discovery tool and how it can be employed to discover new knowledge in the form of relationships or patterns emerging from the analysis of darknet advertisements.
Strategic and investigative analysis are intrinsically interconnected. Law Enforcement Agencies (LEAs) need existing tools to be adapted in order to produce new insights from all the data gathered during their investigations, which could later be converted into useful knowledge.
There are a myriad Data Mining and Knowledge Discovery techniques that can be used in order to analyse data in an automatic or supervised way to help LEAs in such analysis and search processes. However, many of the techniques need an enormous effort to first, understand the tool, and second, to imagine how it can be employed with their own data to generate new information. Additionally, supervised techniques need extra effort to feed the system in order to adapt the outputs and obtain meaningful results.
One of the objectives of the COPKIT project is the elaboration of a toolkit in order to collect, extract, analyse, discover, assess and predict by means of suitable tools to enable the injection of knowledge in all steps of the investigation process. This is one of its strong contributions which covers the design and implementation of several methods adapted to LEAs’ necessities.
In particular, COPKIT employs several Knowledge Discovery (KD) methods as a first step in the exploration of collected data that can come from COPKIT LEA partners’ data or from an environmental scanner. The data can be securely handled by the Secure Test Lab tool also developed in COPKIT in order to test the different KD methods in a safe way.
The information obtained by the different KD methods can be later used by LEAs to refine the research process, focusing on those parts revealed as novel or interesting and also as an enriching piece of knowledge in later steps of the investigation.
Knowledge Discovery tool to find patterns in darknet advertisements
Knowledge Discovery tools can be employed in different ways in order to discover, analyse, assess or predict emerging threats in the criminal field.
Particularly interesting are the non-supervised methods that enable automatic data exploration by obtaining co-occurrence patterns in a data collection. In this regard, COPKIT has developed a methodology to obtain interesting knowledge, in the form of patterns, in a set of collected advertisements by means of a web scraper.
The KD tool developed has been applied to find new insights in the firearms trafficking domain, offering information in the form of IF-THEN rules, with frequency and reliability values, about:
- Temporal trends in advertisements selling specific types of weapons.
- Connections between two or more entities that can be, for instance, two suspects, two cases.
- Grouping/relating entities: for instance, relating similar users of the dark web forums selling a specific type of arm.
The information obtained can be later used by LEAs incorporating it in their investigation, focusing on those relations revealed as novel or interesting to analyse in more detail.