Tracking and preventing radicalisation using an AI 'spy'
TA suite of smart software being developed by EU-funded researchers is designed to help law enforcers and intelligence agencies stop violent online radicalisation in its tracks.
© #96393538 | Author: theaphotography, 2018 fotolia.com
Terrorism is an urgent threat. In 2015-16 alone, more than 20 terrorist attacks occurred in the EU. Every terrorist involved was radicalised by propaganda and recruited via social media, and researchers believe that, with the right tools, they can use data mining to help prevent future attacks.
Before the internet, the radicalisation process could take years. Now, it could take just a few weeks or months. Detecting cyber propaganda early is crucial, but social media platforms are not equipped to address the issues and trawl through the billions of messages sent every day through Twitter, Facebook and closed channels such as Telegram and WhatsApp. That is where the EU-funded INSIKT project comes in.
Even larger agencies have only a few people assigned to watching social media all day long, explains project coordinator Jennifer Woodard of Insikt Intelligence in Spain. It varies across Europe, but they are mostly using existing tools for social media listening purposes that arent really built for looking for this type of stuff. We can give law-enforcement agencies a platform built for their very specific needs.
AI to counter a changing threat
INSIKT researchers are taking forward a unique data-mining platform, designed to detect vulnerable individuals and identify online radicalisation as it happens.
The back-end technology for their INVISO intelligence software has already been proved fit for purpose by identifying the Twitter user who posted a viral video of a radical imams speech promoting violent action. It has also been used to identify bank transactions funding terrorist activity, and to decipher the coded discussions inside a WhatsApp drug-dealing ring.
EU funding is now enabling researchers to develop the front-end platform with input from European law-enforcement agencies.
With a pilot programme under way, law-enforcement agencies should have access to a user-friendly platform that offers accurate, multilingual and real-time detection of jihadization by the time the project closes in 2019. The intelligence software should allow law enforcers to limit the spread of radical content, to detect when propaganda is being released and to intervene early.
The system uses a combination of tools to study the dynamics of radicalisation and the potential for violent acts within social media platforms.
Natural language processing allows the system to understand written languages. Sophisticated text-mining algorithms sift through enormous amounts of textual data to identify radical content, suspicious messages and covert radicalisation processes.
Social-network analysis maps and measures the flow of information between people and groups to identify radicalisers and their relative importance within their community.
Machine learning enables the system to interpret and learn from new data without the need for human input, equipping it with the ability to adapt to changing methods of online radicalisation, while deep learning means the system will learn to automatically develop new ways of working. These models could also be used to detect other kinds of serious criminal activity.
A terrorism-fighting tool
The INSIKT team is working with three European law-enforcement agencies to design the front-end software for the INVISO Intelligence software and turn the system into a commercial product.
The idea is we start with them developing the software to fit their requirements and understanding how it fits into their work flow, says Woodard. Towards the end of the project, theyll be testing a fully deployed system which will be fit for purpose for law enforcement and intelligence agencies of all different types.