In a Crime Ridden Dystopia We Can’t Afford to be Cautious with A.I. by Adrian Timberlake

Some alt text

Taking down a kingpin in an organised crime gang or disrupting a terrorist plot may involve a number of operatives performing covert surveillance on subjects of interest (SOIs), including following the SOI to monitor their actions. It goes without saying that covert operations – bearing in mind that intelligence agencies are targeting some of the most dangerous criminal suspects, including those with involvement in terror plots or serious and violent crime – can sometimes pose substantial risk to the operatives involved. Intelligence and law enforcement agencies commonly use technology to streamline operations, perform surveillance on SOIs, and to gather intelligence.

What technology is used in covert surveillance?

Many people appear to worry that new AI-powered technology, such as facial recognition cameras, may invade privacy, yet almost every one of us carries a unique tracking device on our person: our mobile phones.

Mobile phones are commonly used in the planning and execution of organised crime and terrorism, but data capture networks that can intercept 2G, 3G or 4G transmission are commonly used to track and intercept communications between those plotting serious crimes. Now that AI will increasingly be used to police online communications and social media, to discover online crime, we will hopefully see a future where criminals have no place left to hide.

Smile for the camera

Cameras are a common surveillance tool as they can meet a number of public security goals: enhancing security over a wide area, capturing evidential-standard imagery (high-quality images that can be used to aid prosecution in a court of law) and deterring crime. We’re all familiar with CCTV cameras, but the optics used in intelligence and law-enforcement operations are of a much higher quality, out of necessity, and need to meet certain requirements, such as providing clear, identifying images to operatives in poor light, darkness or fog. Electro-optical/infrared systems (EO/IR) in cameras are commonly used; this is a combination of electronic and thermal imaging. The combination of EO/IR systems means that operatives will be aware of both what is visible, but at too great a distance to spot with the naked eye, and what is hidden – ie. people inside buildings.

EO/IR sensor systems are often paired with a target point tracker, a capability achieved by the combination of many different algorithms. The target is initially located by a human operator and, once the target has been acquired, a feedback control loop will adjust the equipment to keep the target in the sensor’s field of view. This means that the operative does not have to constantly adjust the position of the equipment and risk missing a potential threat or the capture of evidence-gathering imagery. Some camera systems also have the ability to follow GPS tracking devices placed on targets.

Short-range EO/IR cameras are commonly used for civil or tight surveillance purposes – to provide security for a building or to conduct surveillance on an individual, for example. They benefit from a compact size – which means they can be moved and deployed by one operative – and provide evidential-standard images.

Effectively, short-range EO/IR cameras trade surveillance over a wider area for higher-quality images. Although EO/IR equipment is developing constantly, presently, the wider the range in which the camera can detect movement or threats, the lower the quality of the images. However, in some scenarios, this is a trade worth making. Long-range EO/IR cameras are commonly used in hostile environments in which situational awareness is a higher priority than evidence-gathering. This would include military operations, for example, in which awareness of potential threats, such as approaching vehicles and convoys, is key to an operation’s success and the safety of personnel.

Medium-range EO/IR cameras are commonly used in law enforcement, border control and counter-terror surveillance, as they compromise a balance of providing situational awareness to personnel and the ability to clearly see potential threats at a distance from which it may be necessary for personnel to act.

How could AI augment these systems?

There is value, from a situational awareness and planning POV, of obtaining all available information as quickly as possible on potential new threats. It would be more useful to operatives, who must decide how to act, to have an approaching person identified against a database of SOIs and to be notified of any potential threats, rather than just having the ability to see a person approaching from a distance. Facial recognition that uses AI is one of the most powerful identifier systems that exists today, and can be implemented into many models of surveillance equipment.

Using AI-powered technology may mean less operatives are needed on the ground

Intelligence and law-enforcement agencies are doing an incredible job at protecting national security and keeping the public safe with the resources and technology available to them. Unfortunately, a lack of resources means that operatives cannot be everywhere at once or watch every suspect 24/7.

MI5 has around 3,000 active subjects of interest at any one time and lacks the resources to put a large number of people under surveillance at the same time. Keeping a SOI under 24-hour surveillance, in cases where the SOI is considered to pose a high-level of risk to public security and safety, requires a whole team of officers.

AI could help agencies to save valuable time and resources. Facial recognition, for example, could assist agencies in tracking SOIs, their locations and activities, when personnel cannot be spared to conduct covert surveillance on them. In addition, facial recognition software can be programmed to automatically alert relevant law enforcement personnel if a SOI comes within the range of the camera – this may be useful for enhanced protection of buildings or areas suspected to be the target of a terror plot.

Facial recognition may also be able to compensate for human error. Targeted facial recognition – which is the type of facial recognition commonly used by law enforcement agencies – identifies faces by comparing biometric measurements, such as bone structure and the distance between a person’s eyes, to a database (called a ‘watchlist’) of images and biometric measurement profiles of a number of people wanted by the police. The camera, in this case, will only be looking for matches to faces in its database. The idea behind the use of biometric measurements is that if a suspect were to change their appearance slightly, or use easily available disguises such as wigs and coloured contact lenses, the camera should still be able to recognise them, even if a human might not.

For years, facial occlusion – the alteration of facial appearance by sunglasses, scarves and masks – has affected facial recognition performance to differing degrees, but now facial recognition developers all over the globe are tackling this issue, and tech companies in China have managed to build facial recognition that can identify people wearing masks. Although facial recognition technology has been subjected to criticism over accuracy, its capabilities and accuracy are improving at a rapid rate.

The need for powerful technology that frees up resources trickles down into civil policing as well. With the current threat of rising knife crime, which is creating a culture of worry and fear in communities and leading to children carrying knives ‘for protection’, the slow implementation of technology that uses AI in policing, and high amount of public backlash in instances of police forces implementing new technologies, is problematic. Law enforcement organisations in high-tech countries including the U.S. , China and the U.K. are increasingly making use of facial recognition with the aim of identifying and apprehending known suspects in serious and violent crimes. The London Metropolitan Police recently deployed live facial recognition to enhance public safety and security on the streets of London – the move was both supported and criticised, and led to the arrest of a woman wanted by police for a violent offence. Only time will tell if this is enough to convince the British public that AI does have the potential to enhance public safety and security

Adrian Timberlake

This image has an empty alt attribute; its file name is Adrian-Timberlake.jpg

Building on nearly 20 years as a Scientific Officer in the Ministry of Defence, Adrian is now the Chief technical director of Seven Technologies Group and specialist in tech that uses artificial intelligence (AI) for law enforcement and counter-terror operations.