A Captcha is a form of Test of Turing making it possible to differentiate in an automated way a human user of a Ordinateur.

Because the test is carried out by a computer, in opposition with the standard tests of Turing realized by the human ones, Captcha is often described like a test of reversed Turing. This term is nevertheless ambiguous because it could as mean as the participants as try to prove as they are computers.

Applications

This test is used on Internet in the forms to guard against the automated and intensive tenders carried out by malevolent robots .

The checking uses the capacity of analysis of Image or its of the human being. Usual Captcha requires thus that the user types the letters and the visible Chiffre S on a Image distorted which appears with the screen. Some Web sites prefer to post an image which contains a mathematical question.

They are used:

  • against the Spam:
    • at the time of the inscription to free Webmail S (of which the accounts could be used thereafter for the sending of not requested mails),
    • during the tender of messages in discussion forums and Blog S (which could make it possible to make Spamdexing), etc;
  • against the automated extraction of databases;
  • against the attempts at Attack by rough force;
  • for the participation in surveys (whose results could be distorted by automated votes).

Concerning the name

“Captcha” is the acronym of the English “Completely Automated Public Turing test to Such Computers and Humans Apart” , that is to say in French, “public test of completely automatic Turing having for goal to differentiate the human ones from the computers”. This term, which is a registered trademark by the Université Carnegie Mellon, was invented in 2000 by Luis von Ahn, Manuel Blum and Nicholas J. Hopper of this university, and by John Langford of IBM.

History

At the beginning of Internet, the users always wanted to make the text illegible by the computers. The first were the Hacker S , posting on sensitive topics in forums on line, which were automatically supervised with keywords. To circumvent these filters, these hackers started to replace the words by characters visually resembling. For example, HELLO could be replaced by |-|3|_|_ () or) - (3££0, as well as a multitude of other numerical alternatives. Thus the keyword filters could not obviously all detect them. This process was later known under the name of " 13375p34k" (Leetspeak).

The first reflection on the creation of automatic tests which could distinguish the human ones from the computers with an aim of controlling the access to the services Web appeared in a manuscript of Moni Naor of the Institut of science of Weizmann, gone back to 1996, and entitled Verification off has human in the loop, gold Identification via the Turing Test . Of Captcha primitive seem to be developed later, in 1997 at AltaVista by Andrei Broder and his colleagues with an aim of preventing robots from adding sites to their Search engine.

By seeking a means of making their images resistant to attacks of software of character recognition, the team sought in the handbook of their digitizer of mark Brother , which gave recommendations to improve the performances of the Character recognition (standard of similar writings, melts homogeneous…). The team designed puzzles while trying to simulate what could cause a bad automatic recognition of characters. In 2000, von Ahn and Blum developed and published the concept of Captcha, which included/understood any program which could differentiate human from a computer. They invented multiple examples of Captcha, of which first Captcha which was largely used (by Yahoo! in particular).

Characteristics

Captchas, by definition, are entirely automated, requiring only one small human intervention during the use of the test. This thus presents obvious benefit to the level of the costs and performances.

The algorithm used to create Captcha is often public, although it can be Brevet E. This is done with an aim of showing that to break this type of test the resolution of a difficult problem in the field of the artificial intelligence requires, rather than simply the discovery of the secrecies of the algorithm, which could be obtained by decompilation or other means.

Accessibility

The tests of Captcha based on a reading of text - or any other task of visual perception - make impossible the access to the resources protected for visual defective people . Nevertheless, Captcha inevitably does not need to be visual. Any problem of Artificial intelligence, like the Voice recognition, can be used as bases for a test of Captcha. Certain implementations of Captchas make it possible to the users to choose audio Captcha.

The development of audio Captchas seems to be late compared to the visual tests. Other types of tests, as those which require a comprehension of text (for example, a logical puzzle, questions, or instructions to create a Password) can also constitute methods usable within the framework of Captcha. Once again, there are only few studies concerning their resistance vis-a-vis the countermeasures.

Some interesting tests appeared with the idea of the recognition of images. KittenAuth is a test of this type, which requires of the user to recognize an animal (kittens) in a series of photographs of various species (dolphins, pups, foxes, etc).

For the visual defective people (like the users blind man S, or having difficulties with the perception of the colors), Captchas visual present serious problems. Owing to the fact that Captchas are conceived not to be not read by the machines, the current tools of assistance as the readers of screen cannot interpret them. Owing to the fact that certain sites can use the tests of Captcha as of the initial process of inscription, or even with each connection, the latter can completely block the access. In certain jurisdictions, the owners of sites can become the target of litigations if they use of Captchas which discriminate people having certain handicaps. In other cases, those which have visual difficulties can choose to identify a word which is read to them.

Although to provide audio Captcha allows the blind users to read the text, this process always excludes the people who are at the same time defective visual and auditive.

The use of Captcha thus prevents a great number of individuals from using all the services based on Internet like PayPal, Gmail, Orkut, Yahoo!, as well as many forums and systems of blogs…

This being known as, even for perfectly conspicuous people, the new generations of Captcha, conceived to resist the sophisticated software of recognition, can become practically impossible to read. Even certain examples present on the sites quoted below are indecipherable with the eyes of the majority, not to say totality, of the human ones.

A report/ratio of W 3 C.A. underlined the inaccessibility of certain visual tests antirobots.

Skirting

There are several approaches to put in failure Captcha:

  • one can use a human labor to recognize them;
  • to exploit the bugs in the implementations which make it possible the attacker to completely pass in addition to Captcha;
  • to improve the software of character recognition.

Human labor

It is possible to pass through the test of Captcha by using human operators employed to decode Captcha. A publication of W3C indicates that such an operator “could easily check hundreds of Captchas per hour”. Nevertheless, this solution is not inevitably economically viable. Spam die however succeeded in circumventing the difficulty by creating Internet sites which require so that the user passes a test of Captcha to reach it, this test being does of it that necessary by another site, such that of Yahoo to validate the creation of a new address email. The user of the first site contributes thus, without his knowledge, the malevolent acts of the latter. A countermeasure exists: to add, in the captcha, an expression identifying its transmitter clearly (such as “yahoo.fr”).

Bugs of design

Certain protective systems by Captcha badly conceived can sometimes be forced without using software of character recognition, but simply re-using the ID of a session of a known image of Captcha. Sometimes, if part of the software which generates Captcha is located side customer (the validation is made on a waiter, but the text which the user must seize to be identified is generated side customer), then the users can modify the software customers to post the text of Captcha not deformed for example.

Automatic recognition of characters

Although Captcha were initially conceived to counter the software of standard character recognition used for digitalization by scanning of documents, several research projects proved that it is possible “to overcome” a great number of Captchas with programs specifically adapted to a type of Captcha. For of Captchas with deformed letters, the adapted approach is generally made up by the following stages:

  1. removal of the bottom of the image, for example with filters of colors and the detection of fine lines;
  2. segmentation, i.e. cutting of the image in several segments containing only one letter;
  3. identification of the letter contained in each segment.

Random links:Chaumeil | Private education in France | Climate of Green Bay (Wisconsin) | Jean-Louis Matterel de Saint-Maixent | Just of Trieste