AI-powered tools to keep up with online extremists

Dr Brenton Cooper at Fivecast explains the critical role of advanced open-source intelligence solutions

 

The pivotal role of open-source intelligence (OSINT) in helping to foil extremist bomb attacks at a huge Lady Gaga concert in Brazil recently is a signpost to the future for all organisations engaged in public protection.

 

Online extremists calling themselves “Little Monsters” and motivated by anti-LGBTQ beliefs sought to inspire others to use improvised explosives against the 2.1 million-strong crowd on the beach at Rio de Janeiro. The alarm was raised by Rio de Janeiro state police intelligence after they identified the group’s online activity, leading to the issuance of search warrants and the subsequent seizure of electronic devices, followed by arrests.

 

The group allegedly operated through online digital cells. It used violent and self-harm-related content, coded language and extremist symbols in a plan to build an online community of teenagers. It was the ability to detect this content and then explore what the group was doing on social media – and presumably other channels – that enabled the plot to be nipped in the bud so effectively.

 

Violent content: an extremist recruitment tool

Unfortunately, all around the world, the use of violent, graphic content to lure young people to extremist causes online has become a distinct trend. Police authorities and intelligence agencies are struggling to keep up with the volume and speed at which this material emerges. Last year, for example, Vicki Evans, senior national co-ordinator for counter-terrorism and Deputy Assistant Commissioner at the Metropolitan Police highlighted the problem of children as young as 10 being subjected to a “pick and mix of horror” through a variety of social media.

 

Depictions of extreme violence, often without any ideological motivation, are growing in popularity, as youngsters have easy access to content promoting misogyny and mass violence. Evans admitted that it took counter-terrorism officers a large amount of time to find and examine such material.

 

Online radicalisation: targeting younger audiences

The problem of online radicalisation of young people is serious. Since 2018, six of the nine deadliest mass shooting events have been perpetrated by young men – not the middle-aged men more usually associated with such crimes prior to the mass use of the internet.

 

Online content has the ability to recruit and desensitise young people, persuading them to subscribe to conspiracy theories and extreme beliefs. They can connect with fellow extremists almost anywhere, giving them a sense of belonging and validation, as has clearly been the case with solo attackers inspired by Islamic radicalism or far-right causes. 

 

Scale of online material is too great for humans

The scale of material is the most immediate difficulty for the authorities. The statistics website Statista, for example, estimates 5.42 billion people will use social media this year, generating masses of data. The mainstream platform, TikTok, achieved five billion downloads alone last year.

 

Many platforms lack genuine moderation and have algorithms that help spread content, making it problematic for authorities to determine whether those posting malicious content are part of a wider group and what their aims are.

 

The sprawling nature of the internet means intelligence teams must cast their net wide and deep to include niche platforms such as 4chan, 8chan/8kun, Discord, Telegram, TikTok, WeChat and Weibo. The “chan” sites are often home to various far-right communities, providing anonymity for users to share hate speech and radical beliefs. Many groups operate more extensively on dark web forums, which are not indexed by conventional search engines and therefore will only surface in specialist browsers.

 

The task of analysing this information from fast-emerging, often-amorphous extremist groups is substantial and often beyond the embryonic OSINT capabilities of policing or counter-terrorism analysts. Few policing organisations have the budget for an army of fully trained open-source data analysts.

 

Harnessing AI for OSINT analysis

To filter and analyse such a mass of data at speed, organisations must be ready to adopt purpose-built technology. While human skill must always be paramount, the power of AI vastly augments the tradecraft of experienced investigators.

 

AI-powered OSINT tools save analysts days of manual data analysis, detecting suspicious content in relation to set risk parameters. Intelligence teams can customise risk detectors to their specific intelligence requirements, using analytics to establish connections between individuals and groups associated with extremist content.

 

Combined with other forms of human and traditional classified intelligence, OSINT technology provides vital insights that teams cannot extract unaided. Analysis goes beyond keywords to assist in examining images, videos, memes, posts and other multi-media content types. It is now possible to detect text in images through optical character recognition technology and train detectors with the power of machine learning to find specific objects, concepts and text meaningful to investigations.

 

Keeping pace with extremist innovation

Advances in OSINT algorithms also enable agencies to keep pace with the changes in the meaning of online jargon or symbols as threat groups adopt them. If the technology cannot reveal the main entity posting content, associates with online links can often be identified. Policing intelligence teams using an OSINT platform can also store video from multiple accounts ready for viewing when required, which is an effective tool when groups post content and quickly remove it in a bid to outsmart detection.

 

AI, when applied to OSINT, highlights not only content but also suspicious patterns of behaviour and gives insight into levels of engagement and probable goals. Smart prioritisation removes what is irrelevant, with augmented intelligence enabling intelligence officers to make better and faster decisions.

 

Once investigators are on the trail of a plot or believe online extremists are actively seeking to recruit, the technology can expand its searches more deeply to collate risk-indicators from diverse online data sources including the deeper recesses of the dark web, criminal marketplaces, people databases, official watch lists and curated datasets of known threat groups.

 

As highly malign and motivated groups become skilled in the use of content, and more young people become exposed to extreme video and images, policing organisations need access to AI-powered OSINT technology.

 

By vastly augmenting their own intelligence skills and experience, they stand to benefit hugely from automation and algorithms developed by the defence and intelligence communities to analyse online content.

 

If policing, intelligence and counter-terrorism agencies are to protect large numbers of innocent people from the consequences of online radicalisation or extremism, they are going to need AI-powered OSINT technology to augment their own skills. 

 

---

 

Dr Brenton Cooper is CEO and Co-founder of Fivecast

 

Main image courtesy of iStockPhoto.com and stevanovicigor