Balancing Security & Privacy with Facial Detection Technology

The use of biometric markers for security became commonplace almost ten years ago when Apple introduced Touch ID. Now we could more securely lock and unlock our phones, purchase apps and log into our online bank accounts with a scan of our fingerprint. Today, biometric identification has expanded beyond the fingerprint scanner on our phones to facial recognition technology in video security systems. But the proliferation of video monitoring in our everyday lives has led to increased concerns around privacy rights and biometric bias.

Research has shown that facial recognition systems misidentify people of color far more than white people. In 2020, a Michigan man was arrested for a crime he didn’t commit because of a faulty facial recognition algorithm. Concerns around bias and abuse grew so heated last summer that IBM stopped all facial recognition development, and Amazon and Microsoft said they would stop selling their facial recognition technology to U.S. police departments for one year.

Alternatively, facial detection technology — or person detection — is much less controversial than facial recognition, but it can have significant security and analytic value when used in the right setting. These two technologies are similar, yet vastly different in their privacy implications. It depends on which tool you need for the job. Facial and/or person detection provides exactly the kind of information required for many occupancy analytics, for example, but it does so without identifying individual persons. This makes it far more palatable to many organizations, and circumvents the growing number of state and local laws that prohibit or restrict facial recognition technologies.

Facial Detection vs. Facial Recognition

Facial detection is the process of locating a human face in a video or image. If you’ve ever used a face filter on social media or Zoom, that technology uses facial detection to identify where to place the filter over the face. Facial detection only focuses on finding the face amongst other objects, not on identifying the individual to whom it belongs.

Facial recognition technology not only detects that there is a human face in an image but also goes a step further to analyze the image and then identify whose face it is from within whatever database is at its disposal. This is done algorithmically by comparing the image to a database of stored records. The technology is used at airports to securely confirm that your face matches your passport photo, by federal and local law enforcement to identify criminals, or even to identify victims of abuse.

In order to work properly, the algorithm has to be trained with a robust dataset of millions of images of what the algorithm needs to recognize. In the case of facial recognition, it would be millions of images of faces. The more faces you feed the algorithm, the better it gets at identifying physical characteristics. The privacy issue here is clearly that you can be passively identified without your awareness or consent, which many of us find objectionable. Facial detection, of course, has no such implications.

The Value of Facial & Human Detection in Security

Many industries use facial and person detection technology to gain useful insights while maintaining individual privacy. Retail stores use person detection technology to understand the flow of traffic in a store, or whether someone visited a fitting room and then abandoned the purchase. Companies like Uncanny Vision can even measure shopper intention based on a person’s body language while in the store, like if they spend time looking at an item or pick it up off the shelf. This data helps retailers optimize their store layouts for better flow and point out areas where they could increase sales of specific items.

This concept can also be applied to larger spaces, such as conference centers and event spaces. AI-driven detection technology can identify the entrance and exits with the highest foot traffic, areas where people linger, and how many people are in a given space at one time. In the age of coronavirus, this is even more important. Controlling the flow of foot traffic in a physical space while maintaining ample social distance is vital, and facial detection technology can help do that. Last summer, Amazon implemented their distance assistance software that detected when workers were not socially distancing in the warehouse.

As the world slowly begins to get back to the office, facial detection technology can assist with monitoring health and safety in the workspace. Touchless thermal temperature scanners use facial detection AI to guide a person to position their face in the correct spot on the screen in order to accurately measure their body temperature. That same technology can also identify whether that person is wearing a mask and remind them to do so if they are not. When combined with a larger access control system, this data can either approve or deny access to a building based on body temperature and face mask use.

As demonstrated in these examples, for general building security and property management, facial recognition is not necessary to keep a building secure or control the flow of foot traffic. When someone gains unapproved access to a building or an area of a building, it’s more important that the surveillance technology identifies that it’s a person causing the security breach, not that it can identify the individual by name. While the security threats are still real and we must do everything we can to protect against and prevent them, facial detection technology can be a part of keeping buildings safe while avoiding individual privacy concerns.

Facial recognition still has its place, especially in circumstances of implied consent, such as frictionless identity authentication at the workplace or anywhere else you have agreed to make this exchange of convenience for privacy.

As is so often the case, it’s a matter of picking the right tool for the job.