Fingerprint-Protected USB Sticks Cracked

juct writes "Manufacturers of USB sticks and cards with fingerprint readers promise us that their data safes can only be opened with the right fingerprint. In their tests, heise Security found that it is easy to bypass the authentication and get access to the protected data. This works by sending a single USB command, using the open source tool PLscsi, that changes the accessible partition. They found the vulnerability in several USB sticks that use the same chipset. The article concludes: 'The fingerprint sensors in the products mentioned above apparently only serve one purpose: they mislead interested buyers. They do not provide any significant level of protection. We can only recommend that these products not be purchased.'"

Comments

Mythbusters

[TheMeuge]

Didn't Mythbusters beat a bunch of fingerprint readers a couple of seasons ago? I seem to recall them using printed pictures of fingerprints with great success.

http://www.youtube.com/watch?v=oXyFmieZjiE

Comment removed

[account_deleted]

Comment removed based on user account deletion

Hardware-based security is often vulnerable

[Lucas123]

Corsair's Flash Padlock has the same issue. You can open the case through a single screw in the back of the drive and then access an electronic switch on the board, which can be easily tripped with a piece of wire, giving you access to the memory chip without having to punch in a security PIN. Hardware security methods just aren't as secure as software-based encryption.

Re:Hardware-based security is often vulnerable

[mpapet]

Hardware security methods just aren't as secure as software-based encryption.

You couldn't be more wrong about biometric authentication. You probably haven't seen the Sagem or Cogent sensors implemented well. It is the very rare organization who would actually spend the money to do the job right. A revision is necessary to make your statement accurate.

Cheap and dirty hardware security methods just aren't as secure as software-based encryption.

That's better.

Fingerprint scanners suck.

[mpapet]

It goes without saying that there are a large number of low-end sensors disguised as excellent front-ends to biometric authentication. You need to segregate two things.

1. the sensor itself.

2. the implementation of the sensor. (e.g. sensor as a front end)

There are two legitimate sensor manufacturers in the U.S. and one very well-known French company all of whom do not sell to just anyone anywhere and at prices absolutely out of range for a TV show and the average company.

Another thing to keep in mind is even IF there was budget for a good device, (oh to dream) there are implementation issues that can make the hardware worthless. As is often the case, meaningful implementations tend to complicate practically all business/operations matters which is why no company bothers.

To generalize that all fingerprint scanners suck is just wrong.

Re:Fingerprint scanners suck.

[u8i9o0]

But no two scans of one fingerprint are identical pixel for pixel. If you scan one thumb ten times, you get ten different hashes.

Then that's not the way it should be done. For one thing, while the angle of the print may change, the relative size will not.

I think you can create fingerprints based off of a formula. All you need is to supply a set of variable coefficients. The hash would be that set of coefficients for your formula.

It's been a very long time since I had studied fingerprints, and that was rather cursory.

From what I know, every print has at least one point. The alternative is that some prints have ridges going straight across, which doesn't sound right to me.

- Focus on the most prominent one or the one ranked highest in priority.
- Measure the distances between unique points and their angles relative to each other.
- A left loop will always be a left loop no matter the rotation, and has an apex.
- Same with a tented arch, except it will also have a triangular shape.
- A whorl has two epicenters of a given distance.

I never worked in the field, but the above plan seems obvious to me. I also don't have a large sample set to help refine that formula - maybe having two whorls or two similar loops or some other combo never happens.

With any authentication, the important thing is that it be easy to produce the key and make it very hard to fake it. Therefore, the biggest problem with fingerprint authentication is that the user keeps leaving their key everywhere they touch. It's like mentioning your passwords in plaintext within every conversation you have. One solution may be to use toeprints instead.