As spam assists in the creation of artificial intelligence
October 16, 2008, 8:22 am
As a user of various Internet resources, each of us has repeatedly passed the test CAPTCHA (for England. Completely Automated Public Turing test to tell Computers and Humans Apart - completely automated public Turing test to tell computers and humans), which is to introduce a few characters, presented in graphical form in the special field. Such a challenge, albeit a very simple man, is extremely difficult and almost impossible for the different bots. We know that CAPTCHA tests now do the task kind of a filter to separate the living person from the program for sending spam messages by registering free accounts to e-mail-services, leaving messages on the forums, in comments to blog posts, etc. On the one hand, to date, this decision is one of the most effective ways to fight spam and spammers, but on the other - absolutely not ruled out that it is a test CAPTCHA, strangely enough, will give impetus to improve the artificial intelligence. After all, CAPTCHA, in essence, is nothing but an option Turing test, and if the boat is able to successfully address of its objectives, we can safely claim that we are representative of artificial intelligence - similar to programs that compete for a prize Lebnera . And most importantly, that certain progress in this direction already. At the beginning of this year, Jeff Yang (Jeff Yan), a researcher from the University of Newcastle, United Kingdom, presented a program to cope with the test CAPTCHA, used for protection services Hotmail, MSN, and Windows Live to the effectiveness of 60%. Of course, the effectiveness of character a person is much higher and up close to 100%, but do not forget the fact that per unit of time the computer is able to solve with varying degrees of success, thousands of tests. In other words, the program Jeff Jan protected with a CAPTCHA ineffective. the sake of fairness it should be noted that the algorithm, written by Ian, sharpened by the decision of tests CAPTCHA, used only in case of Hotmail, MSN, and Windows Live, and programmer not to reveal all secrets until such time as Microsoft does not optimize security system. In other words, the program was able to detect with high efficiency characters in some form, while other variations remained for her CAPTCHA intractable. Nevertheless, it is understandable that once written the first algorithm for solving the Turing test, it will soon emerge and better solutions - and wrong - all the same Jeff Yang at the upcoming conference on security ACM Computer and Communication Security Conference threatened to show more a clever program that is capable to solve much more different tests CAPTCHA. No less interesting is the offspring of Philip Gaulle (Philippe Golle) from the Research Center in Paolo Alto. Written his boat specializes in pattern recognition animals - also variation test CAPTCHA - and in most cases correctly defines the contours of a cat or dog. Of special interest program is due to the fact that recognize the image of an animal much more difficult than defining characters. And even the increased complexity of the problem does not affect the outcome - of the 100 images correctly recognized more than 83. Another case that the testing is necessary to process 12 images, thereby reducing the likelihood of a successful solution of the problem until 10, 3%. immediately raised fears that these algorithms will be used spammers to increase the quantity and quality of its attacks on services, forums and blogs, but accurate data on bringing artificial intelligence to pass filter in the form of CAPTCHA tests yet. But this does not mean that in the future, as growth in computing capabilities of computers, intelligent program does not do tests CAPTCHA amusing anachronism, to add a simple user problems, rather than protect them from spammers.
captcha, test, program, tests, intelligence, artificial, spam, jeff, turing, other• Jeff Yan + CAPTCHA
