Facial Recognition Systems: Another Threat to Privacy?

Are you on Facebook? Do you worry about how much Facebook knows about you? Well, as much as it knows now, it’s about to know much more. Facebook has been investing heavily in artificial intelligence (AI) technology to identify your face uniquely and track your behavior more precisely. Facebook’s facial recognition tool, called DeepFace, is nearly as accurate as the human brain in recognizing a face. DeepFace can compare two photos and state with 97.25 percent accuracy whether the photos show the same face. Humans are able to perform the same task with 97.53 percent accuracy. DeepFace was developed by Facebook’s AI research group in Menlo Park, California, and is based on an advanced deep learning neural network. Deep learning looks at a huge body of data, including human faces, and tries to develop a high­level abstraction of a human face by looking for recurring patterns (cheeks, eyebrow, etc.). The DeepFace neural network consists of nine layers of neurons. Its learning process has created 120 million connections (synapses) between those neurons, using four million photos of faces. Once the learning process is complete, every image fed into the system passes through the synapses in a different way, producing a unique fingerprint among the layers of neurons. For example, a neuron might ask whether a particular face has a heavy brow. If so, one synapse would be followed; if no, another path would be taken. DeepFace soon will be ready for commercial use, most likely to help Facebook improve the accuracy of Facebook.com’s existing facial recognition capabilities to ensure that every photo of you on Facebook is connected to your account. (Facebook has one of the largest facial databases in the world for its photo tagging service.) DeepFace might also be used for real­world facial tracking, for example to monitor someone’s shopping habits as that person moves from physical store to store. Facebook could profit handsomely from the detailed behavioral tracking data collected through DeepFace. Facebook is one of many organizations using facial recognition systems, and neural networks are one of several techniques for this purpose. The Oregon Department of Motor Vehicles (DMV) uses facial recognition software to ensure that driver licenses, instruction permits, and ID cards are not issued under false names In Pinellas County, Florida, police can capture a 3D video and upload it to an image gallery for comparison to identify people with prior criminal records or outstanding warrants. Whatever the technology foundation, facial recognition systems are raising alarm among privacy advocates, who are worried about the far­reaching use of people’s facial photos without their knowledge or consent. Although police departments and DMVs have strict limits on the use of their facial recognition software, casinos are beginning to faceprint their visitors to identify high rollers to pamper, and some Japanese grocery stores now use face­matching to identify shoplifters. Dr. Joseph J. Atick, one of the pioneers of facial recognition technology, is at the forefront of these concerns. Atick is in favor of facial recognition for specific purposes such as law and immigration enforcement, motor vehicle department authentication, and airport entry, but he warns about its use for mass surveillance. Atick has been encouraging companies to adopt policies that safeguard the retention and reuse of facial data, stipulating that it cannot be matched, shared, or sold without permission. Facial analysis has progressed beyond scrutinizing static features. Frame­by­frame analysis can isolate involuntary millisecond­long expressions, revealing private sentiments. Although these insights can drive productive endeavors, they are fraught with privacy implications. For example, do you want the person conducting your job interview to be able to review a videotape, identify fleeting moments of confusion or indecision, and decide against hiring you? People might try to use the software to determine whether their spouse was lying, police might read the emotions of crowds, or employers might use it to monitor workers or job applicants secretly. In addition to revealing people’s emotions without their consent, these facial recognition tools might misinterpret people’s feelings. Psychologist Paul Eckman studied these fleeting microexpressions that surface when people are  attempting to suppress an emotion and devised the Facial Action Coding System (FACS). Forty­three facial muscles control seven primary expressions—happiness, sadness, fear, anger, disgust, contempt, and surprise. Combinations of other basic muscle movements signal more advanced emotions such as frustration and confusion. People, and now computer programs, can be trained to recognize the universal spontaneous micromovements that divulge peoples’ true feelings—narrowed eyelids, raised eyebrows, wrinkled forehead, scrunched nose, flared nostrils, or tensed lips. Ekman has served as an advisor to Emotient, a San Diego startup whose software can recognize emotions from a database of microexpressions that happen in a fraction of a second. Emotient has worked with Honda Motor Co. and Procter & Gamble Co. to gauge people’s emotions as they try out products. Emotient is confident that the ability to gauge customer emotions objectively and accurately will give retailers more tools to increase sales. Ekman says he is torn between the potential power of all this data and the need to ensure that it is used responsibly without infringing on personal privacy. Although facial expression analysis will likely never be an exact science, academics, business people, and, certainly, government agencies are intrigued by its possible applications. Online learning could be improved using webcams that perceive confusion in a student’s expression and trigger additional tutoring sessions. Cameras could sense when a trucker is exhausted and prevent him from falling asleep at the wheel. When voice and gesture analysis and gaze tracking can be combined with facial expression analysis, the possibilities will explode, along with the privacy implications.

 

Case Study Question

1. What are some of the benefits of using facial recognition technology? Describe some current and future applications of this technology.

2. How does facial recognition technology threaten the protection of individual privacy? Give several examples.

3. Would you like DeepFace to track your activities on Facebook and in the physical world? Why or why not?