Reversing a Fingerprint Reader Protocol (2021) | Hacker News Reader

Reversing a Fingerprint Reader Protocol (2021)

blog.th0m.as

65 pointsthejj10010016 comments

As noted in the article I reversed the protocol for a related Goodix device (which was on Intel so used actual SGX instead of the white-box): I used the firmware update system to insert additional vulnerabilities in the sensor firmware and extract the PSK from that side.

I did a talk about it here: https://www.youtube.com/watch?v=IyjUY-xvFw4

Damn, I always thought that the fingerprint data was encoded somehow and never left the sensor hardware itself! OS-level access to the imagery seems like a security risk, but also opens some interesting possibilities for alternative uses.

AFAIK it depends per reader. This one seems to be a weird webcam on steroids, but others do the matching locally.

IIRC, none of them do it particularly securely.

What's the security status of fingerprints on phones? Surely they don't leave the security chip? I hope?

I don't think fingerprints should be regarded as a secret.

Can you please post a link to high quality images of your own fingerprints? It should be fine, probably nobody has the technology to make them show up on a threatening letter mailed to the government, or anything like that.

Of course I won't, but then again I would send you pictures of any other body parts the same.

If someone gets a hand on anything you touched, they have your fingerprint. Last time you traveled to another country - did you have to give them fingerprints? Is the software running your phone closed source? Could you ambush me at night near my house and forcefully take them?

All I am saying is they are so weak as a secret that rhey should not be regarded as one.

AFAIK many phones store the fingerprints on-chip. I haven't looked too deeply into it, though, so it's possible there's a brand out there that streams fingerprint information as a video.

On Android, there are different levels of biometrics: https://source.android.com/docs/security/features/biometric If your fingerprint scanner reports Class 3/STRONG, hardware key stores are a requirement. Anything Class 2/WEAK or higher is supposed to make sure a kernel compromise cannot leak keys/authenticate to the OS. If it's Class 1/CONVENIENCE, simply running the biometrics in the trusted execution environment (think "secure VM acting as TPM") is also permitted.

On iOS the TPM/secure element deals with credentials, they're not submitted to the CPU.

Author here, didn't expect to see this on HN today! If you've got any questions, shoot!

Do you have more posts similar to this one? Noticed your blog was a bit empty...

Ha yeah I should really get on updating some of the info there. Got derailed with work quite a bit.

Most recently did some work on BitLocker: https://news.ycombinator.com/item?id=42747877

> It then proceeds to generate a new, random, PSK and sends it to the device. This represents a trust-on-first-use security model.

Wow, i expect them using hardcoded PSK, with PSK is flashed in factory.

The real work ha underneath the software eg. I can't write a camera driver but thankfully someone else can

That's cool the raw data image GIMP

I didn't follow the byte ordering of the image format at the end. Anyone have an explanation?

You have four 12 bit values, they are packed into 6 bytes. Camera image formats are different but I am guessing this is probably MIPI RAW12?

EDIT: You have the code in the repo. https://github.com/tlambertz/goodix-fingerprint-reversing/bl...

Excellent

Reply on news.ycombinator.com