The Future of Redaction in a World of Facial Recognition Technology

In today’s CaseGuard article, we’ll discuss how facial recognition works, as well as the ethics of using such software. We’ll also cover how the growth of facial recognition technologies will lead to more and more organizations to adopt redaction software.

Facial Recognition and Privacy Concerns

As recently as January 15, 2020, the Future of Privacy Forum (FPF), a Washington-DC based thinktank, had their Senior Counsel and Director of AI and Ethics, Brenda Leong, testify before the House Committee on Oversight Reform regarding the privacy and ethical implications of facial recognition technology. Leong’s testimony outlined how the use of facial recognition technology has caused concern among privacy advocates, policymakers, and the public.

Their testimony could hardly be more timely. Today, facial recognition technology is used in surveillance systems, fraud prevention systems, check-in systems at airports, and even in medical monitoring. As facial recognition technology’s use becomes more pervasive, many organizations are wondering how to balance convenience against privacy rights when adopting the technology.

To effectively understand and evaluate the risks involved in facial recognition technologies, businesses need to be aware of how facial recognition technology works and what type of data it is collecting. As Leong noted in her testimony, consent over the collection, storage, and interpretation of data is at the core of this debate.

Facial Recognition and Biometric Tracking Technology

Facial recognition is a form of biometric tracking technology. Biometric tracking technology makes it easier to both identify and authenticate individuals quickly and reliably through their unique biological characteristics. Essentially:

• Authentication is a process of comparing data to stored data or a template of the person’s individual characteristics to determine if there is a match. Authentication answers the question, “Are you X?”
• Identification answers the question, “Who are you?” In the process of identification, the person’s characteristics, facial features, fingerprint, or other data are compared to the contents of a database to determine the person’s identity.

Facial recognition is one of many forms of biometric recognition, including iris recognition, fingerprint identification, and gait recognition.

How Facial Recognition Works

There is some confusion regarding facial recognition technology. Not every camera system uses facial recognition, and there are different types of image identification systems that are often incorrectly lumped in with facial recognition. These other types of systems are not necessarily connected with individual identification. In fact. some merely detect the presence of faces. Among experts in the field, there are generally considered to be five levels of identification within surveillance or facial recognition technologies. Each level has a distinct purpose ranging from object detection (i.e., this is a face) to identifying an individual’s face.

Here are the five levels of identification with facial recognition technologies:

1. Detection

This is where the technology determines if there’s a face in the picture. It is commonly used to count customers in surveillance footage or even organize photo albums. At this level of identification, there are few privacy concerns, as no object detected has been uniquely identified.

2. Characterization

This second level of identification determines if the object it’s tracking is a human face. It then can answer specific questions, such as whether the face it sees is a man or woman. This type of recognition technology can be used to display gender-specific ads or even track customer shopping habits in-store.

3. Unique Fixed Identifier

During this phase, the technology detects and characterizes the image. However, it does not directly tie that information to any known information on the individual. A common use for this is to track a customer’s shopping experience, patterns, and other in-store behaviors. There is some risk to privacy if the data is connected to other identifiable information.

4. Verification 1 to 1

This level wants to know whom the system is looking at. It asks the question, is this person whom they say they are? These can be used for a variety of uses such, including as an identification tool at an ATM. Since the person is identifiable, there is a risk to privacy. There should be privacy notices with express consent given. Security breaches of any type can lead to the loss of personally identifiable information (PII).

5. Identification 1 to Many

This is the ultimate form of identification. However, even the best algorithms can’t guarantee that identification will occur. It asks the question, “Who is this person?” There are multiple uses available for this level of identification including photo-tagging, tracking consumer loyalty, and conducting hyper targeted advertising. Due to the high risk of abuse of privacy, steps must be taken to secure the data.

New Uses of Facial Recognition Technology

By 2024, it’s estimated that the global market for facial recognition technologies will create over $7 billion in revenue. While new uses and applications are created with the features of facial recognition, the most significant area of growth is predicted to be in surveillance in the public sector. That being said, analysts predict the use of new technologies in facial recognition are expected to be in three particular use cases:

• Law Enforcement or Security: the main advantages are to increase the ability to combat crime and terrorism. In some cases, facial recognition has been used to identify suspects in connection with the investigation of violent crimes.
• Healthcare: machine learning algorithms and facial analysis will make it more and more possible to track patient’s medication use, detect genetic disease, and support pain management therapies.
• Retail Marketing: Know Your Customer (KYC) marketing is an emerging field. An example is a recently designed data system by Facebook, which, while analyzing shopper behavior, pulls information from their social media profiles to assist sales staff.

Redaction Software

With the growth of facial recognition technologies, there will also be considerable growth in the area of redaction software. Companies who use facial recognition technologies at all levels will need to protect the privacy of the individuals’ data that they collect. There is currently precedence for new laws and policies to balance surveillance technologies with individual personal privacy. While companies enjoy the new markets and security that facial recognition brings, they will inevitably have to comply with privacy laws and protect the individual and keep their data secure.

One way to do this is with video redaction software. Since many companies store data for an expressed amount of time, prior to data storage or even sharing of data, using redaction software can eliminate the privacy risk to the individual. An easy-to-use redaction system, like CaseGuard, uses automation and machine-learning to redact faces and other identifying information with ease. If the identifying information is completely removed prior to data storage or sharing information with other companies, then many privacy risks to individuals can be safely elimination.