How secure is Face ID?
Apple’s description of enrollment and comparison is very similar to Touch ID. The enrollment sends data through a one-way channel to the Secure Enclave, a special tamper-resistant chip bound deeply inside the iPhone and iPad architecture that can only respond with limited information, such as confirming a match was made when unlocking for Apple Pay and the like. Secure Enclave also stores some other private information.
As a result, Apple doesn’t collect this information and process it centrally, nor does it store it on the device in a manner that can be retrieved by cracking a phone, a phone backup, or intercepting information to and from it.
However, the concern remains that, with proprietary technology under the control of Apple, a government could force changes that would pass or extract facial identification information, or perform comparisons with faces that a government is looking for.
In the current hardware architecture, however, that seems unlikely. Apple has engineered its systems so that there’s no reasonable way to rework it to change the flow of facial (or, with Touch ID, fingerprint) information to a different source. It would have to create a whole new kind of phone and new firmware.
Apple notes that developers can create encryption keys protected by Face ID that are stored in an iPhone X’s Secure Enclave. These keys can be used by the developer’s app entirely within the Secure Enclave, performing operations that are as protected as Apple Pay and biometric identification. It will be interesting to see how developers make us of this additional security level. (It doesn’t seem as if this is an option for Touch ID.)
Can someone fool Face ID?
The answer should be no, based on the approach Apple has taken: number of data points, use of infrared scanning, and attention awareness. With other systems, people have used photographs, plaster models, and other approaches that Face ID would seemingly resist. A plaster model doesn’t offer the same reflection as a 100 percent identical human face, because infrared reflects off living skin differently than off an inanimate material.
Apple says in its white paper that it introduces randomness to make it even more difficult. The sequence of 2D infrared scans and depth-map dot captures are sent in a random order, and the project dot-pattern is both random and unique for each device. This makes it harder for an attacker to use predictable elements to dupe a scanner—or they might succeed in fooling an iPhone X in their possession, but fail with any other iPhone X out there.
The Face ID security white paper notes in passing that while there’s a primary neural network that performs identification, a second neural network checks against spoofing, looking out for telltales of photos and masks.
Sign up for Computerworld eNewsletters.