As promised I am back after two weeks with a load of information about biometrics and related topics. We have everything from billion dollar lawsuits to issuing toilet paper using face. Enjoy!
Unlocking your phone with eyes closed
New Google phone can apparently be opened even if your eyes are closed (e.g. while sleeping), which is not great (slightly easier to exploit than using your finger while you are asleep).
Apple requires you to look at the screen, not sure how that works with sunglasses, but you cannot have everything :).
Google claims it can take months to be fixed, which is strange, because open eyes detection is already solved problem (at least for basic security).
Unlocking your phone using any finger
Unlocking your phone while you are asleep sounds bad, unlocking it with any fingerprint is even worse. That is however what happened with new Samsung phone and its in screen sensor. All you need is clear silicone phone case over the top of the sensor.
Samsung seems to be doing better than Google (anyway, Google problem is not that huge in comparison) and a fix was already provided, although some banks suspended fingerprint authentications on that phone in the meanwhile.
Using facial recognition to save toilet paper
China has introduced facial recognition to toilets to prevent people from stealing or overusing toilet paper. You get 2 feet of toilet paper and that’s it. I don’t know what you are suppose to do, if that is not enough.
Facebook facing massive lawsuit
It seems like the Facebook could be ordered to pay up to 35 billion dollars to users from Illinois based on BIPA regulation for using facial recognition on users from this state without their consent.
They will of course appeal the supreme court.
Facebook tries to block facial recognition in video
Facebook has developed an encoder and decoder for video that blocks facial algorithms from identifying the faces in the stream. This is nice but I don’t know if Facebook tries to help people or just protect their own monopoly on facial recognition in Facebook photos and videos (and at the same time look good for the courts e.g. see above).
New approach to re-identification
University of Surrey have developed a deep neural network that could help improve accuracy levels of facial recognition, especially re-identification cases in which people are appearing in front of the camera short time after first detection.
OSNet is able to identify many small details from a suspect, such as logo on a t-shirt or the type of coat worn by the suspect.
Minnesota new bill about drones and facial recognition
Minnesota senate has specified cases in which police will be allowed to used automated drone with facial recognition, including “Conducting a threat assessment in anticipation of a specific event”. The rules are pretty broad.
Make money spoofing liveness check
FaceTec is offering up to 30k to if you can spoof their ZoOm liveness check system. Ideal opportunity to make some money and help the industry to be better at the same time.
Push-back against digital identities
Human rights organizations are calling for thorough review of digital ID programs. “Each digital identity proposal must be questioned and evaluated to check if it benefits the users, empowers their rights, and effectively protects them from potential risks.”
There is not a lot to disagree with, we should always think about how we apply technology.
Petition to stop facial recognition at festivals
There is an online petition that monitors festivals and wants them to prevent from using facial recognition. Some issues seem valid, some bit overblown, technology itself must always comply with the laws.
More information can be found in the opinion peace on Buzzfeed. While some points are valid and should be considered and regulated, some seem bit strange to me from European point of view, e.g. allowing police to arrest criminals it is looking for.
Why Isn’t Functional Programming the Norm?
Here is an interesting talk about the reasons why functional programming is not the norm. Is it the killer apps? Platforms? Ease of upgrading? Or something else? Find out.
Estimates or #Noestimates?
I understand the origins of #noestimates movement, but always felt that it is more about unwillingness to do estimates right. This article pretty much summarizes my issues with avoiding estimates.
How to explain machine learning to humans
And last but not least, here is an interesting paper about how to do explainable machine learning (if possible) and what does it mean.