11:52 Biometric identification the reliability of the technology | |
| Argued that the biometric identification - it is very reliable and safe thing. I'm not the first year of work in this area and propose to look into this issue in greater detail by the example of human identification by fingerprints. What is it like? Look at my finger. You see a lot of lines, which then converge, then diverge. Here are the points of most Convergence Divergence construct so-called "biometric model." It is all different and daktiloskopisty can almost accurately confirm or refute the involvement of a person for fingerprints. But, first, daktiloskopisty take full hire model thumb, and secondly, in view of all the lines, drawing, and other things. We can follow only part of a scanned fingerprint. Which includes a certain number of points ("main"), but not all. There are so-called (roughly) "threshold", ie, the number of points that must be matched. The higher it is - the more reliable and more difficult to achieve proper matching of prints (identify person). Because the finger, as a result, it is necessary to attach the same or almost the same as on the reference scan. And it is not easy. In addition, as mentioned earlier, we see not the whole model of print, but only part. Accordingly, the check we will also only part of the points. This makes it necessary to "the threshold of sensitivity." And everything is good, and models for people at all different, but the fact is that sometimes we have very similar to prints. "Like two peas in a pod", with slight differences. And here we have a problem. Comparing a person with different people (identifying), we may well be mistaken - a "false alarm". And right to the system confuses the manager of one department to another. But when the system is confusing junior - the developer and the CEO is often sad. Especially for the CEO. "Research" refuse (to do without them). In biometrics is the probabilistic key interest rates (thresholds): false admission of FAR (False Acceptance Rate, «level of personal responsibility system"). Ie, the probability that the system will let a "foreign" person. And false rejection FRR (False Rejection Rate, errors). Ie, the probability that the system not let go of "their" rights. Indicators are very closely interrelated. Bandwidth (1? FAR) is called the specificity, the value of (1? FRR) is called sensitivity. Increasing / decreasing the sensitivity of the system we raise / lower its specificity, and vice versa. The situation is slightly rescues verification. But what it is, and how does it differ from the identity? Identification looks something like this: Smith came to the system, puts his finger, the system will happily report: "Bah! .. Ivanov! Hello, been waiting for! " Verification, in turn, looks like this: Smith came to the system, tells her:" I - Ivanov! ", And puts his finger on. The system happily report: "Ba-aa-ah! .. Ivanov did. Hello, where have you been, then? " In other words, the verification comparison is not with many people (fingerprints), but with one specific person (finger). Then the threshold can be set higher. As the situation improves, if the finger is applied twice. If one fingerprint people are very similar, it is absolutely not a fact that will be similar to another. But technology is a much more expensive. On top of that a lot depends on the scanner itself. Good scanners (from I've seen) - a very cumbersome and very expensive. But they can even be used in fingerprinting. However, they rarely take. Most scanners are "cheap and cheerful" and operate accordingly. That we do not directly touch on some of the possibilities "conditionally open" KFOR that they give us. You can easily replace the imprint of man and not knowing any of his password or his username to log in to your account. In general, in my opinion, biometric identification is not as reliable as it seems. Yes, and the fingerprint is not as reliable at this point, we would like. Especially on large volumes of data (ie, when a large number of users). After all, we need to be compared with a large number of prints and likelihood that the system will fail increases. Verification is definitely safer. An alternative may be the identification of the retina. Anyway, looks like much more reliable. But the equipment is expensive and feel it is not yet possible. Recognition of visual images of works not yet safer than the "footprint". From I've seen: the technology even half a man confused. With the right approach. A biometrics approach implies, in any case, thoughtful and deliberate. I would like to add information about the issues that arise almost always. 1. Experience of cutting the fingers was not. 2. Not looking at it, cut off fingers on most scanners do not work, because scanner responds to the warmth of your finger. There are exceptions, but they are cumbersome and expensive. What we see in movies is frank bewilderment. 3. Cut off my fingers never heated, so to answer "what" I can not. P.S. Thank you for your karma, moved at a blog devoted to this topic (I thought) to become the most imminent. Tell me, if missed. | |
|
| |
| Total comments: 0 | |