Exposed – will AI biometrics leave spies out in the cold?

By Professor Fraser Sampson, former UK Biometrics & Surveillance Camera Commissioner

Arguments against the state using AI-enabled biometrics like facial recognition bounce back and forth, but sometimes the job of government is not just to respect anonymity, but to guarantee it. For some, not being found is more than a constitutional entitlement – it is critical to their safety and that of others. Witnesses in organised crime cases may need ongoing protection, and for the highest-level international intelligence sources that can mean a lifetime undertaking as seen in the backdrop to the report on Operation Kenova published in the UK this week. There are many profound questions from the report into activities of the informer codenamed “Steaknife”, but one that may be overlooked is how far the state will be able to offer anonymity in such cases in the future.

Biometric capabilities like live and retrospective facial recognition are rapidly shrinking our worlds, transforming practical expectations of privacy in our digital lives. While most of the debate revolves around unwelcome state intrusion, in cases involving informants the tables are turned – here it’s the state trying to conceal the true identity of an individual in a world increasingly equipped and determined to expose them.

The reach of open-source surveillance has caught up with high-profile crime figures previously living below the radar. In one case, an Italian mafia boss who had escaped from prison in Rome in 2002, where he was serving a life sentence, was captured in 2022 after detectives spotted him on Google Street View in Spain.

Providing new identities has become much more challenging as advanced surveillance technology becomes part of our everyday lives. Armed with reverse look-up facial matching tools and internet scraping, you can wave images from graduation photographs, amateur theatre performances and park runs under the nose of the algorithm which will race off and return with any matches like an AI-enabled Golden Retriever. Openly available search tools need less and less detail to work with and can geotag locations with little more than a shot of a field or a snap of a coastline.

Regulatory restrictions – albeit patchy – on the host state using such tools won’t stop hostile states and the basic capability is within reach of the rest of us if we are improperly motivated to use it.

Advanced technology continually offers us more ways of keeping an eye on things from afar: vehicles, valuables, vulnerable relatives. Even golf balls and pets can be tagged and tracked for pennies – more cool surveillance presents will find their way into stockings again this year. Smart doorbells and fine location apps give us potent remote surveillance capabilities that were once only available to state intelligence agencies. Our private use of sophisticated technology has ‘socialised’ surveillance, completely changing the surveillance relationship between the citizen and the state. Now the intrusive technology is in our hands which is why appealing for dashcam or doorbell footage has become de rigeur in police responses to every incident. The precise power of this aggregated capability has yet to be understood but there is no doubting its impact.

In espionage cases one government will be trying to provide anonymity and another government trying to breach it. In this scenario, technology consumerism has gifted the hostile state limitless ways of keeping tabs on us. Finding a needle in a haystack has changed from simile of the impossible to simplicity itself and AI-enabled biometrics excel at the hunt. The state’s surveillance reach is coming less from police cameras and more from the fact that we are all capturing everything. Obsessive uploading of our every moment and movement has dissolved the boundaries of state surveillance in a way unforetold by Orwellian fiction. Uncounted and uncountable billions of images and other data reference points are being continually captured, shared and re-shared across the globe. Once upon a time this data needed a human to join the dots. Sifting it was a key resourcing decision and who was going to live long enough to get through even a fraction of it? Then along came Machine Learning with recurrent neural networks that could alchemise what was digital chaff into surveillance gold.

Of course, you can still cover up, alter your appearance and hide your features. Despite concern at the UK government’s proposals to create a criminal offence of concealing your face in public, there are very few restrictions on doing so in many countries. Your bigger problem will be maintaining a digital disguise. Beyond face, other biometrics are being added to our menu, from lipreading sunglasses to Wi-Fi routers that can detect people through walls, gait analytics and pulse sensors that are already measuring patterns as individualised as a fingerprint. Paradoxically, going dark on social media is now seen by friends and relatives as a sign that you need to be sought. There’s nothing so high profile as a low profile. Then there is the Streisand effect whereby trying to prevent attention to an aspect of your private life exponentially increases online interest in it.

The arrival of AI-driven surveillance technology and our insatiable adoption of it makes privacy almost indefinable let alone attainable. At the same time the advanced biometric tools required for authentication of true identity to carry out basic activities like banking, digital payments and registering for services are specifically designed to unmask impersonation. Anyone still dreaming of going off grid will soon find that the grid doesn’t have many off ramps these days.

As British intelligence agency GCHQ reveals its Christmas code-crackers, our compulsion to see and share everything is bringing near total digital visibility, turning us into indirect surveillance assets, not just for our own government, but for everyone else’s too. That represents a fatal threat to the traditional tradecraft of asset-handling and undermines the state’s ability to discharge any duty of care it has to the people involved. We are all spies now.

About the author

Fraser Sampson, former UK Biometrics & Surveillance Camera Commissioner, is Professor of Governance and National Security at CENTRIC (Centre for Excellence in Terrorism, Resilience, Intelligence & Organised Crime Research) and a non-executive director at Facewatch.

Article Topics

biometric matching  |  biometrics  |  digital ID  |  facial recognition  |  Fraser Sampson  |  video surveillance