Not that long ago, the ability of an algorithm to recognize your face may have seemed like sci-fi fodder. Today, it’s a reality for social media giants like Facebook to digital camera auto-focus. The term facial recognition encompasses a broad spectrum of technology, from face detection, which as the name suggests, detects a face and allows a camera to focus on it (auto-focus) to actual identification.
How it Works
Facial recognition is the technology driving the auto-focus of a digital camera, as it recognizes a face and suggests a focus. However, the ability to analyze the captured image for classification or identification has expanded the application of the technology. At the next level, facial recognition technology can classify a face into a broader group, for example, a female. This is applied by marketers for targeting specific groups of people to see an ad.
Now, facial recognition can identify an individual. Using biometric identification, the same process used for fingerprint matching or voice recognition, a system imputes data images, analyzes them, and tries to match the new entry to an existing entry in the database. Facial recognition takes it a step further and detects, creates a “face print,” and then can verify or identify, according to an article in The Conversation.
Facial recognition technology is often driven by computer software and can be applied to an existing software system – like a security camera system in a mall, for example. Because people do not always present themselves with the goal of having a face print taken, the software also has to re-size and adjust the captured image before it can generate the print. And, accessories like large sunglasses or hats, or hairstyle changes could make this more challenging.
How it’s Applied
Facial recognition technology is applied in more ways than catching shoplifters at the mall or helping a digital camera capture the perfect portrait. In fact, the technology is used in both the private and public sector, assisting the FBI through its facial-recognition database, which has an estimated one-third of Americans profiled, according to the Electronic Frontier Foundation. The FBI database’s estimated 85% accuracy pales in comparison to private sector giants Google and Facebook, estimated at 99.63% and 97.25% respectively, reports The Washington Post. (To be fair, the photos uploaded to Google and Facebook are thought to be easier to work with). Through Google’s FaceNet and Facebook’s DeepFace, users are given suggestions for people to tag when a new photograph is uploaded. To do this, the algorithm must create a profile, or face print, of a user to enable the suggestion.
Apple’s new Animoji is another new use for facial recognition technology.
Security has also adopted facial recognition technology. The newly launched iPhone X allows users to unlock their phone via a face print; in China, a user’s face can also act as a pass to security offices and authorize ATM withdrawals. This year, the LPGA debuted the technology at their ANA Inspiration tournament in March, scanning spectators and media at the front entrance, and granting media professionals access to the media center through the match. Airports have also incorporated this technology to support screening.
This technology comes with a bevy of concerns, with privacy at the forefront. The slew of available data could eliminate any possibility of privacy, as people could be identified anywhere, and images can be captured remotely and subtly. As the technology improves, it may be easier for this kind of unwanted identification at a micro level as well. For example, a new app allows a user to use a photo to find a person’s social media accounts – which might be handy for connecting with new friends, but concerning if the attention is unwanted. A recent study intended to be published in the Journal of Personality and Social Psychology curated thousands of photos from an online dating site and mapped the data into a model intended to predict sexual orientation. This raises serious privacy concerns for a user who does not want his or her sexual orientation disclosed.
Consent, of course, goes hand and hand with privacy. When Facebook launched its technology, critics believed it should have required explicit consent through an opt-in, as opposed to auto-opting in and giving users the option to opt-out. Stores that feature security cameras often post such a disclaimer so people are aware, but posting a similar disclaimer if the technology was applied to the street wouldn’t be as easy.
As with any evolving landscape, controlling and regulatory this frontier is a challenge. Some have even raised concerns about potential privacy issues of well-intentioned approaches, like Amber Alerts. Police often include a photograph of the missing child when the issue an alert; however, without oversight, there is no control on whether or not the photo is recorded in a facial recognition system. Staying current on an evolving landscape is difficult, and Congress last addressed consumer privacy legislation in 2009. In both Illinois and Texas, people must give informed consent before technology can identify them.
While facial recognition technology brings security opportunities, such as eliminating long security lines and offering the ability to reliably spot “bad guys,” the potential for bias and prejudice still remains. Racial profiling could remain an issue, and governments could use the technology to suspend protestors. And even if the intention is good, the algorithms themselves could have higher identification rates of a specific gender or race, inherently introducing bias in legal or security systems. For example, research from the Institute of Electrical and Electronics Engineers indicates that facial-recognition systems disproportionately misidentify African Americans, which means innocent civilians may be flagged as suspects in crimes when facial recognition technology is applied through surveillance cameras and software.
Innovation in facial recognition technology doesn’t appear to be stopping anytime soon. A professor of computer simulation, Lyndon Smith, from the University of the West of England is developing a system that may replace train tickets, using scans to recognize facial features and validate the users. It could, he believes, replace the PIN system in banks as well. Marketing is another key arena likely to be impacted by this technology. While Microsoft has already developed a billboard that can adjust to show to specific demographics, eventually a shopper’s face could act as their loyalty card, bringing with it historical shopping data like preferences and purchase history, and allowing marketers and retailers to fully customize the advertisements they encounter. Face it: this technology is here to stay.