Facial recognition software (FRS) works with hardware such as surveillance cameras to give you biometric matches to faces drawn from a database of photos and video stills. In this article, we will explore the different uses of facial recognition for security purposes and examine the ethics of this technology.
Biometric facial recognition refers to the measurement of facial characteristics to identify certain individuals. The technology, available on many surveillance cameras with artificial intelligence, enables private security personnel and law enforcement agencies to determine the best response to a potential threat. Although facial recognition has been around for decades, recent advances in AI have made it possible to deploy security systems with facial recognition at its core on a broader scale.
As Wired puts it, security apparatuses that rely on facial biometrics, “include video analytics [that] analyze footage to detect abnormal activities and suspicious persons that pose a threat to an organization’s security.” Facial recognition software “learns” who and what is normal in different situations. FRS identifies persons and behaviors that human operators may miss. As a self-learning application, your facial recognition software increases its intelligence and becomes more sophisticated over time in proactively identifying potential threats.
When it comes to the ethics of facial recognition to identify persons of interest, we need to remember that it is up to us to adjudicate the matches as they register in your system. The technology itself doesn’t have a bias when it comes to matching faces to the images in your database.
According to a 2021 report from the National Institute of Technology and Standards (NIST), facial recognition algorithms for most FRS programs have an error rate of just 0.08%, which means that the ethical problems with facial recognition mostly come down to user misuse.
For instance, the Chinese telecommunications giant Huawei used facial recognition technology to track Uighurs, the country’s minority Muslim population, who have been subject to government crackdowns and imprisonment. Huawei inputted certain facial characteristics to make this profiling possible. Again, it’s the user of technology, and not the technology itself, that poses our most serious ethical dilemmas.
Facial recognition technology has three main purposes: investigative face matching, real-time analysis for increased situational awareness, and authentication for access control.
As mentioned above, face-matching has been around for a long time. Takeo Kanade, one of the pioneers in face recognition, demonstrated in 1970 that computer programs could calculate the distance ratio of anatomical features such as the human chin. FRS has made exponential technological gains in the decades since, especially in the past few years.
Face matching typically takes place in the cloud. Your AI-equipped cameras capture images of people and automatically search the thousands of photos in your database gallery of images and video stills. This process still requires considerable human intervention. Face-matching software does not produce an exact match. Rather it generates probability estimates, giving your security personnel photos to examine for a potential match.
Using facial recognition for investigative purposes has helped security teams track persons of interest in public buildings, stadiums, theme parks, or across an entire city. For retail businesses, face-matching biometrics helps to identify an individual’s consumer tendencies. As a boost to corporate revenue, facial recognition technology can help answer questions: why did a customer visit a faraway vendor in a stadium rather than the one right next to them?
As a tool for business intelligence and investigative purposes, facial recognition software enhances productivity through its focus on consumer behavior, in addition to its main purpose of delivering a safer and more secure workplace (or city).
Many of today’s facial recognition and visual identification software and systems are capable of providing real-time intelligence. Assessing threats and neutralizing incidents as they arise plays a significant role in assisting your security personnel. The main difference between traditional facial recognition technologies and real-time facial recognition is that the latter requires a great deal more computing power and sophistication.
Real-time facial recognition works largely without human intervention. It nonetheless requires that those making key security decisions exercise caution and restraint. While facial recognition is accurate for the most part, its accuracy is only tested in controlled settings. For instance, at large events such as concerts, even the most cutting-edge AI face-matching technologies generate false positives.
While false positives do occur, and your security or IT personnel should always be aware of this issue by providing your system consistent feedback, real-time facial recognition can be a useful tool in preventing harm. This technology helps private security teams achieve better situational awareness. When you know who and what you’re looking for, you rarely enter the scene of a critical event unprepared.
Biometric facial authentication is used for access control management. Facial authentication is different from the above uses of this technology, as individual persons, employees of an organization for example, freely and willingly submit to have themselves identified by AI.
Accuracy is less of an issue when it comes to facial authentication for access to buildings or other sensitive areas. Access control systems outfitted with biometric AI scan collect a matrix of facial data points to improve recognition and speed. This type of access control system is the most accurate, the least prone to error, and has the added benefit of serving as a hands-free solution to provide access to your facilities.
Concerns over the ethical use of facial recognition mainly consider its accuracy, which leads to issues of bias and possible racial profiling. You may have read headlines about Amazon placing a moratorium on the use of Rekognition by police, its premier facial recognition software, after studies showed that other similar artificial intelligence software was less adept at picking out dark female faces compared to white male faces.
Importantly, newer studies have not shown the same degree of racial bias as some of the earlier ones because facial recognition AI has become significantly more advanced and accurate. Another study released by NIST in 2021 shows that facial recognition software is equally right (or equally wrong) across all demographic groups.
In particular, this study found that:
Amazon has nonetheless extended its law enforcement moratorium until Congress passes laws pertaining to the safe and ethical use of facial recognition. As we stated earlier, the technology has reached a point where it no longer appears racial profiling is an issue. This is not to say that those concerned with racial bias should not be concerned. As we saw with the example of Huawei intentionally profiling its Uighur population, the ethics of using facial recognition software in any capacity is only as strong as the ethics of its users.
Any biometric tool, including facial recognition software, is not private, leading to issues of misrepresentation. For instance, it is possible that individuals could create 3D masks of other persons to fool AI.
It is important to note that the matching process in facial biometric data is not statistical — people never present their faces to your AI surveillance camera in the exact same way. The images captured may also vary depending on the time of day, whether the person is wearing cosmetics, and the clothes they are wearing.
As John Day, an AI ethicist, states: “Companies that deploy facial recognition need to be aware people’s faces are the key to their identities. You don’t have a passcode on your facial features. Identity theft is an issue with facial recognition that I think about frequently.”
Deepfakes, which are digitally altered videos where a person masquerades as someone else, brought public attention to the issue of facial misrepresentation through technological means. Imagine if another person committed a crime while wearing a mask of your face. Companies creating facial recognition AI are well aware of spoofing and have begun designing facial recognition AI so that surveillance cameras don’t generate false positives — two years ago, researchers discovered that all it took to fool facial recognition software at the U.S. borders were paper masks.
But those of us in the business of supplying organizations with facial recognition AI now have a better and smarter FRS for wider commercial deployment.
Facial recognition software is a powerful technological tool for public and private organizations to locate, track, and neutralize individuals bent on doing harm to businesses or the general public. It can be used for investigative purposes by private security teams and public agencies. It also increases situational awareness and helps safety and security teams understand what needs to be done during an emergency or critical event. Finally, FRS can be used for access control, providing biometric authentication for organizations looking to limit access to secure areas in their buildings or facilities.
At the same time, facial recognition technology poses ethical questions that we need to consider. Eliminating racial bias will go a long way in addressing many of the ethical concerns of this technology. When it comes to privacy, many states are enacting laws that limit the use of facial recognition. Vermont, for instance, banned FRS without a warrant, except in the case of child exploitation and sex trafficking.
Illinois, Massachusetts, New York, and other states have begun to consider limits to the use of biometric data by private companies. Regulations enacted at both the state and federal levels will have a major impact on the use of facial recognition AI for both public and commercial uses.
Nonetheless, FRS is far too sophisticated at this stage not to develop into an important tool for both public and private organizations, law enforcement and private security, despite potential state and local interventions.