Monday, December 9, 2013

Scientific Computing: Computational Science on Biometric Authentication

What Is Biometric Authentication?
The definition I like the most for biometric authentication is "you are your key." You are authenticating yourself by something you are such as your fingerprints and voice. In this post, I am going to introduce Two most popular biometric authentication methods fingerprints and iris scan, and how scientists apply computational science on them.

Fingerprints
Fingerprint was severed a purpose as a form of signature at in history. However, it has been used more and more as a scientific form of identification in modern society. Today, they are routinely used for identification, particularly in criminal cases.
How does it work?
First of all, we first capture an image of the fingerprint. Then, we enhance the image by using some image-processing tools, and identify and extract some critical points from the enlarged picture.


To authenticate, we use the extracted points to compared with the claimed user's stored information, which was captured and recorded in the past. With some predetermined level of confidence, the computer then determine if it is a match.


For determining a match, different countries have different standard. For example, in England, fingerprints must match in sixteen points while in the United States, surprisingly no fixed number of points are required to match.


Iris Scan


Iris scan is, in theory, one the best for authentication. One reason is that it has little or no generic influence on the iris pattern so it remains accurate for identical twins. Furthermore, the property of the pattern is pretty stable throughout a lifetime.
How does it work?
Iris scan system requires sophisticated software and equipment. First, we use an automated iris scanner locates the iris, and then we get a photo with white and black color. The resulting image is then transformed into a 2048-bit iris code.
Two iris codes are compared based on the formula below:

d(x,y) = number of non-match bits/number of bits compared

For example, d(0010, 0101) = 3/4 and d(101111,101001) = 1/3. A perfect match is d(x,y) = 0 but of course we can't expect perfect match in practice. The expected distance for the same iris is 0.08 while the expected distance for different irises is 0.05.

To Conclude
Compare with password, biometrics clearly have more advantages. In particular, it is hard to forge. However, the more sophistical the biometric system is, the more it costs. That is why people nowadays still prefer password because it is way more cheaper.

2 comments:

  1. I like your post about Biometric authentication. We see it in a lot of spy movies where both good guys and bad guys use it to allow access only to the most pivotal systems. I think that says a lot when we start using it for our cellphones. Whebn we think about it cell phones are the epicenter of our lives. We rely on them to communicate with each other through messaging,calls, and the internet. It does have it flaws though. being able to lift someones finger print isn't as hard as someone might think. or tricking the scanner into thinking its reading a thumbprint thats not actually present. I know the new Iphone's just came out with Biometric authentication and they have had some troubles. That is not to say that one other form or authentication is better, they all have their weaknesses. I just think the technology needs to be more refined.

    ReplyDelete
  2. It's really cool to see features like this applied to the real world. Actually where I work, to get access into the data center we need to first use a fingerprint scanner and then and iris scanner. I think you explained the fingerprint reader and the iris scanners perfectly. There are issues with biometric scans as they are possible to forge. The biggest problem is that while a bad key or password can be revoked, it's not clear how to revoke a bad biometric.

    ReplyDelete