When most of us think about what the future of online security is we think of biometrics. Whether it is being able to scan your thumb to open your front door or using facial recognition to log into your computer the future is nearly here. It is the next great leap in cyber security and in crime fighting but it is not quite ready to be rolled out for everyone, not just yet anyway.


Amazon has designed a facial recognition system that could help to pave the way towards making our technology more secure. Of course any system like that needs to be tested so it was. The system, known as Rekognition, was tested by the American Civil Liberties Union (ACLU) using the 535 members of Congress and comparing them to 25,000 public mugshots. No members of Congress were included in the mugshots (you can insert a joke here about deserving to be if you would like to) but 28 Congressmen or women were matched up falsely.

Now all jokes aside about every member of Congress being a criminal it should be evident that the system is not ready for real world use. A false match could have serious implications in real world usage. The wrong person could be identified and possibly sent to jail for a crime they did not commit because the system erroneously matched them to someone else or the wrong person was able to gain entry to a place or something that they should not be in.

Why the test failed

Amazon stated that the test was done using the default setting of 80% match and blamed the results on poor calibration. They recommend that law enforcement use the setting of 95% and law enforcement has been trialing the software in several locations around the country. It is available through Amazon Web Services and costs less than $12 per month for an entire police department so one can certainly see the appeal with tight municipal budgets.

Facial Recognition

Facial recognition takes into account several different factors like the shapes of the eyes, nose and mouth, the space between each and any bone structure that is visible and compares it to others in the database. What has been learned through experience the ACLU claims is that African-Americans and women tend to produce higher error rates as highlighted by 11 of the 28 false matches coming from members of the Congressional Black Caucus. The test performed has prompted several members of Congress to request the number of law enforcement agencies using the software and to ask what safeguards are in place from Amazon CEO Jeff Bezos.

As far as law enforcement is concerned facial recognition software is becoming a key tool in their belt. It is already used all over the globe, from London to Beijing and it has proven to be both very effective (Chinese police use it to track and apprehend criminals) and ineffective (the London Metropolitan Police have found it is 98% inaccurate). Facial recognition today is a far cry from what it was in the 1960s when it was first developed and in many cases it is easily able to outperform humans when matching photos. The software is even becoming mainstream as it is used on social media platforms like Facebook and Snapchat.

There have been successes with facial recognition software already. Criminals have been tracked and arrested much more easily than using standard police tactics in several cities all over the world. In Mexico it has been used to prevent voter fraud and it has been used in some US states to prevent people from obtaining a false driver’s license. It has also been used to secure our technology like smartphones and in some cases even our money. It is the future, it is just not perfect.

Comments are closed.

Scroll to Top