One of the more exciting announcements during Google’s Pixel 8 unveiling was the return of a secure version of Face Unlock. But, after eleven days of testing the phone, it’s clear that the technology isn’t ready to challenge Apple’s Face ID yet.
Google debuted 3D facial scanning biometric security in its 2019 Pixel 4 range, which was a first for Android. It worked through several sensors, the most important being a flood illuminator and a dot projector. The hardware and software combined to create a hyper secure way of verifying your identity. Google now says it is able to achieve the same level of security through “machine learning”.
After nearly two weeks of trying it out, that appears to be true. I have signed into secure apps like password vaults and paid for things with my face. The problem is that it can be hit and miss when it comes to successfully verifying my identity in certain conditions.
In good lighting, it’s fast and accurate. In low light, however, the technology rarely works. I was testing how the Pixel 8 Pro camera handles a mixture of skin tones in dark a bar with friends at the weekend (stay tuned for my review) and Face Unlock didn’t successfully do its job once.
I had similar issues outside when my face was illuminated by evening streetlights, although it was better. At home, with the lights down, it struggles too. The good news is that you can use Face Unlock and the fingerprint reader simultaneously, so if one fails you can switch to the other.
Dedicated facial scanning hardware in the iPhone 15 and Pixel 4, like the flood illuminator and dot projector, means they work in low light conditions. Both are insanely fast and accurate too. It would be great if Google offered more details about precisely how machine learning works when it comes to Face Unlock (I’ve asked) and if this experience can be improved at all with updates.
It is, at least, secure. Google says the Pixel 8’s AI-powered Face Unlock has the strongest level of Android biometic security (Class 3). That means the system has a low spoof acceptance rate (SAR) of less than 7% and a false acceptance rate (FAR) of 1 in 50,000. So someone with a picture of your face, or someone who looks like you, has virtually no chance of fooling the Pixel 8 into unlocking.
A recent post on Reddit about a Pixel 8 owner’s brother tricking the system isn’t something I was able to repeat with my brother. I also held up printed pictures of myself, my brother and a very similar looking cousin to the phone’s camera, but it never faltered.
We don’t know the conditions of the Redditor’s test. Perhaps the Pixel 8 was tilted just enough to capture an image of the verified user’s face, which I have managed to do at tight angles. Or this was the one in 50,000 occurrence that the system can be beaten. Or maybe it’s just made up. But in my testing I haven’t been able to scam Google’s Face Unlock technology yet.
It’s impressive that Google has managed to achieve a higher level of security with just the selfie camera and some software trickery, but there’s work to be done here because it struggles in low light. I don’t know if that can be fixed with a software update. But, more likely, if the company values this feature, we might see a dot projector and flood illuminator next to the selfie camera in the Pixel 9.
Follow me on Facebook, Threads and Instagram
Read the full article here
