Government agencies around the world have started experimenting with using facial recognition technology in their surveillance and security efforts. While this surveillance may seem concerning, perhaps even more scary is that marketers are also using facial recognition data for identifying and tracking their customers.
Companies spent $3.35 billion on facial recognition technology in 2016. That number is expected to grow to $7.76 billion by 2022. Organizations are constantly finding new applications for face recognition and new ways to use the facial data they’ve already collected.
- The History of Facial Recognition
- What is Facial Recognition?
- How Does Facial Recognition Work?
- Who uses Face Recognition Technology?
- How Accurate Is Facial Recognition?
- Face Recognition AI is Biased
- How to Beat Facial Recognition
- How to Turn Off Facial Recognition on Facebook
- Which Stores Have Facial Recognition?
- Airlines Use Facial Recognition
The History of Facial Recognition
Facial recognition technology has been around for decades. The concept was first used in the 1960s by Woody Bledsoe, Helen Chan and Charles Bisson. Its original application was nowhere close to as advanced as today’s face recognition technology. The original systems would collect a set of measurements on a person’s face (pupillary distance, mouth width, distance from chin to forehead, etc.) and compare them to images in a database.
“This recognition problem is made difficult by the great variability in head rotation and tilt, lighting intensity and angle, facial expression, aging, etc. Some other attempts at face recognition by machine have allowed for little or no variability in these quantities. Yet the method of correlation (or pattern matching) of unprocessed optical data, which is often used by some researchers, is certain to fail in cases where the variability is great. In particular, the correlation is very low between two pictures of the same person with two different head rotations.” -Woodrow Bledsoe, 1966
Older versions of face recognition failed to account for simple factors such as lighting and head rotation, but today’s systems like Apple’s Face ID can quickly and accurately identify faces even with different lighting and head angles.
The recent improvements are due in large part to massive improvements in camera technologies. Modern cameras are smaller and deliver higher resolution images of faces, so facial measurements can be more numerous and more complex.
What is Facial Recognition?
Facial recognition is a software based tool that collects facial data and maps an individual’s facial features and stores that data as a “faceprint”. The system then uses a camera to gather facial data and runs their faces against the faceprints to find a match.
Automated facial recognition systems work similarly to how people identify faces. There are key features of a person’s face that let us know that a face belongs to someone we know or a complete stranger. The same distinct features that you may rely on to determine if you know a particular face may also let a computer system know that a given face matches a photo in a database.
How Does Facial Recognition Work?
Cameras capture an image of a face and pick out distinct details about a person’s face and then compare them to other people’s facial recognition data. A person’s face is read as a series of measurements between distinct points on their faces.
Who Uses Face Recognition Technology?
- Airport Security: US Customs and Border Protection have been using facial recognition technology since 2015 to screen non-US residents on international flights. The US government has detailed its plan to expand its use of facial recognition to US citizens whenever they leave the country. It hopes to reduce the reliance on physical documents like passports and paper tickets. TSA currently verifies travelers’ identities by checking photo identification, but it believes an automated system will be more accurate and efficient.
- Smartphone Companies: Smartphone manufacturers, specifically Apple and Samsung, use facial recognition for user authentication to unlock their devices.
- Colleges: Some universities use facial recognition for taking attendance. Universities in China use face recognition to verify student identities for exams to make sure that who is taking the test is the right student.
- Social Media Companies: Facebook uses face recognition for identifying people in digital photos. You may have seen this first hand if Facebook shows you photos that you are in but have not yet been tagged in.
- Brick & Mortar Retailers: Stores have begun using facial recognition to determine which products people are looking at and to gather information about how people shop. Anothear application is to catch shoplifters. Stores may maintain a database of known shoplifters and then check each customer as they enter the store to make sure they aren’t dealing with a shoplifter.
- Marketers & Advertisers: Marketers and advertisers use facial recognition to improve their campaigns’ effectiveness. Facial recognition can give them a clear idea of who they are marketing to, and which messages will be most effective.
How Accurate Is Facial Recognition?
In July 2018, ACLU reported that Amazon’s facial recognition tool “Rekognition” falsely linked members of congress to a database of mugshots. The ACLU tested photos of 535 members of congress against a database of 25,000 publicly available arrest photos. While none of the members of congress appeared in any of these arrest photos, the system found 28 matches.
There are two types of errors that apply to face recognition: false positives and false negatives. False positives are when a recognition system matches a person’s face to a facial profile in the database, but that match is incorrect. False negatives, on the other hand, are when a system does not find a match in the database for the person’s face when there is in fact a profile matching the subject in the database.
As with most security systems, there is a trade-off with facial recognition of security and convenience. A more secure system may deliver a few false negatives when trying to identify you, before ultimately giving you access. A faster system may give you access faster, but the focus on speed could give access to unauthorized users, if the system mistakenly find a false positive.
According to the National Institute of Standards and Technology (NIST), between 2014 and 2018, face recognition software got 20 times better at searching a database to find a matching photograph. Facial recognition systems failed 5 percent of the time in 2010, 4 percent of the time in 2014, and just 0.2 percent of the time in 2018. The large improvement from 2014 to 2018 is due in large part to improvements in machine learning capabilities. While most providers have improved their algorithms substantially, there is still a range of accuracy of
Read More: Ongoing Face Recognition Vendor Test Part 2: Identification – NIST
Face Recognition AI is Biased
A report published by Joy Buolamwini of MIT Media Lab and Timnit Gebru of Microsoft Research explained that artificial intelligence are more likely to falsely identify darker skinned individuals and women. This is because the databases feeding the AI are made up primarily of lighter-skinned subjects. As a result, the highest error rates fall on populations that are already subjected to discrimination and biased policing. One of the biggest problems with artificial intelligence is that it often continues and strengthens the biases of the people who created the AI.
How To Beat Facial Recognition
Fighting surveillance can mean tricking facial recognition cameras with special clothing. This clothing uses patterns that mimic human faces, and make it difficult for the camera to find your actual face.
This example looks like digital camo, but is a pattern that confuses the facial recognition cameras that are in use today. It includes pixelated “face-like” patterns that can keep the surveillance system from recognizing your actual face next to the patterned fabric.
As creepy as these look, they are effective in fooling facial recognition systems. If you are trying to completely hide from facial recognition or other surveillance systems. These masks are effective in hiding your true identity but the systems could still recognize the mask and follow you activity, although linking it to a “false identity”.
How to Turn Off Facial Recognition on Facebook
Facebook introduced its facial recognition features in early 2018. You may have noticed Facebook suggesting tags in photos of your friends or yourself.
- Go To Your Profile. Under your profile photo, on the right, select ‘More’.
- In the options that appear, select ‘View Privacy Shortcuts’.
- At the bottom of the page select ‘More Settings’. Then select ‘Face Recognition’.
- Under the explanation of why Facebook uses facial recognition, you should see ‘Do you want Facebook to be able to recognize you in photos and videos?’
- Click on the question and toggle face recognition from ‘Yes’ to ‘No’.
Which Stores Have Facial Recognition?
Many brick-and-mortar retailers have begun using face recognition. In a report from March 26 by the ACLU, many retailers were secretive in their use of face recognition. The ACLU reached out to the top 20 retailers in the US and of these 20 companies only one said that they do not use facial recognition. That one company is Ahold Delhaize, whose brands include the supermarkets Food Lion, Stop & Shop, Giant and Hannaford. Only one company confirmed that it used face recognition: Lowe’s. It says it uses the technology to identify shoplifters.
Large companies with confirmed use of facial recognition:
- Home Depot
It’s unacceptable for stores to use facial recognition to track their customers without giving some sort of notice. Putting a clause on a website is not enough to let customers visiting physical stores know that they are being recorded and identified with surveillance technology. Expecting consumers to check a website for this disclosure before visiting a store in unreasonable. Organizations that are using face recognition right now should put policies in place that protect consumer privacy despite the delay in related laws and regulations. According to Peter Trepp, CEO of facial recognition firm FaceFirst, putting policies in place now will help to encourage public support of the technology and minimize consumer fears.
In China, customers at KFC can pay with facial recognition technology. KFC teamed up with Alibaba to let users of the Alipay app pay for their orders just by smiling, once their face has verified their identity.
Ethics of Face Recognition
Face recognition technology brings about many of the same privacy issues as social media and other internet tracking mechanisms. It’s not obvious to people when their facial data is being recorded, and companies are still very secretive with how they are using people’s faces in their business practices.
“The technology is about saving time but it’s not without some controversy. Republican Sen. Mike Lee and Democrat Ed Markey have called on Customs and Border Protection to stop expanding the biometric program, partially mandated by Congress, until the agency implements privacy regulations and provides a report to Congress on the viability of biometrics.” –CBS News
If every person that enters a store is being recorded and compared to a database of facial recognition data, that information needs to be disclosed at the door. Otherwise people are given no chance to opt out of the blatant surveillance these companies are using.
Airlines Use Facial Recognition To Speed Up Travel
Delta Airlines launched its first biometric airport terminal in November 2018. The system allowed travelers to go from the front door of the airport to the plane without showing a passport. The system photographs each traveler and compares it with a digital copy of travelers’ passport photos stored in a Customs and Border Protection database of people flying that day. The process can take as little as two seconds.
This could make airport travel much smoother and more efficient for most travelers. However, if the biases that we mentioned earlier carry over to these biometric systems, minorities could find that travel becomes a burden, as they will be the ones pulled aside on false positives by the system.
Conclusion: Facial Recognition Could Be Great, Or It Could Be a Complete Distaster
Facial recognition has already proven that it can bring great improvements to currently clunky systems. However, it has also shown that implemented poorly, facial recognition can have a number of negative outcomes. The biggest improvements will likely come in those systems that rely on high-accuracy identity verification and are currently slow due to this need. That means that airport lines, where every person currently has to show form of ID, will get much shorter and travel times will get faster.
The biggest downside we see to facial recognition is that governments can use it to identify protesters, and for broader tracking of their citizens. Use of facial recognition for evil is certainly possible, and it’s likely already happening in places like China. Facial recognition could aid in businesses making big strides in the effectiveness of their marketing or it could lead to a great system for businesses that makes shopping a terrible experience for customers.
The future of facial recognition and its application is still unclear, but it could make huge changes to the business world as a whole. Whether those changes are positive or negative on a wider scale is yet to be seen.