RU2608262C2 - Distortion of symbols for symbolic and graphic reverse turing test - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: invention relates to means for graphic distortion of separate symbols by means of font interpolation (morphing) to each other. This invention consists in fact that downloading two or more fonts, in order to obtain by means of interpolation (morphing) one or more intermediate states between them, and using results to generate images for graphic reverse Turing text.

EFFECT: technical result consists in providing possibility of creating images for completely automated Turing test (CAPTCHA), while increasing complexity of automatic recognition of given symbols using computer programs.

3 cl, 1 tbl

Description

Technical field

The invention relates to a method for graphically distorting individual characters by morphing (transforming) fonts among themselves, used to display text.

Turing's fully automated character-graphic reverse test (SARTSNA) is designed to distinguish between computers and people and is based on offering users an image of a set of characters (usually, the Latin alphabet or Arabic numerals), drawn with distortions. The essence of such a test is that the creation of programs for automatic passing of Turing tests (PAPATT) of such tests is as difficult or impossible as possible. Therefore, SARTSNA developers often offer a wide variety of ways to distort the displayed character set. Nevertheless, this method also has a downside, which is a huge load on a person when passing such tests.

State of the art

Among the analogues of the invention can distinguish such libraries as "Cryptograph?" (Http://www.captcha.fr/) and "Securimage" (http://www.phpcaptcha.org/). They provide ample image generation capabilities for SARTSNA with distorted characters. A comparison of them with the invention is shown in table 1.

Table 1. Name, Description Using an arbitrary set of different fonts Create new fonts based on existing “Cryptograph?”, Http://www.captcha.fr/index.php? Page = 34 Yes No (U "Securimage", https://github.com/dapphp/secur image / blob / master / securimage.php ^* ( ^* unfortunately, the description of the features of "Securimage" is only available as source code) Yes No Invention Yes Yes

They use such methods of distortion of characters as rotation, resizing and shifting relative to each other, scaling, etc. Moreover, the most effective of them are nonlinear, i.e. having time-variable parameters specified by intervals of possible values (for example, a rotation angle of 10 to 30 degrees), because linear distortion i.e. not changing their parameters from image to image, it is easier to compensate when writing PAPATT. It is also worth highlighting the use of specific fonts for character output, as the invention is closely related to this type of complication of automatic passing SARTSNA. It is interesting because it has so far been used on a limited set of available fonts (in most cases 1 non-standard font is used), which means that it allowed PAPATT authors to collect used fonts from available sources or to train their programs (for example, those built on neural networks) with a high degree of probability to recognize the displayed text.

Disclosure of invention

The software result of the invention is the ability to create such image generation algorithms for SARTSNA, the symbols on which will be most understandable to humans, but at the same time difficult for automatic recognition using computer programs. The statement is based on the assumption that if we take 2 human-readable images of text characters familiar to him and find the average between them using some interpolation (morphing) algorithms, the resulting image will be recognized by the person as the same character (1). At the same time, automatic image recognition programs for SARTSNA (for example, based on neural networks) operate with finite alphabets consisting of predetermined symbol images. Therefore, from the point of view of their development, the invention is a non-linear way of distorting the character set. So, its use will have a significant impact on the complexity of developing automatic image recognition programs for SARTSNA and the efficiency of their work, and taking into account statement (1), if many other factors are observed, it will not have a significant impact on human character recognition. In other words, it becomes easier for a person to pass such a SARTSNA, and it is much more difficult or impossible for specially created automatic recognition programs for this.

The invention includes a method for graphically distorting a plurality of symbols that the system depicts for conducting SARTSNA, and a machine-oriented system that implements it. The sequence of actions of the system described in the invention consists of the following steps:

1) Using at least 2 fonts for morphing, and containing images for at least the desired set of characters. For example, you can take 2 fonts: “Times New Roman Regular” and “DeJaVu Extralight”.

2) Carrying out the morphing procedure for each pair of images of the same symbol from different fonts. For example, if there are 2 fonts and 10 characters, then for each of these 10 characters we will take an image from each font to find one or more intermediate states between them by interpolating their points (formula 1); then grouped accordingly, these resulting sets of symbol images will be intermediate states of the originally taken fonts, and they can be used to create new fonts (formula 2).

If necessary, it is possible to create and use key points on downloaded fonts to optimize the morphing procedure. Key points can be noted, for example, angles in raster fonts, if we do not have vector ones.

In addition, it is possible to use instead of 2 fonts, one font and a geometric figure for morphing between them. In this case, statement (1) becomes invalid, since the characters of the font taken will lose their outlines in the process of moving morphing from font to geometric shape. However, to a certain extent, this procedure is also allowed and is also a distortion of characters.

3) We use the resulting set of new images of each character (one or more new fonts) for the images of the characters we need in the SARTSNA images.

Thus, the system generates images for SARTSNA and new fonts based on the ones used. Also, the possibility of using the created fonts to generate others based on them is not excluded, because this further increases the complexity of developing PAPATT.

Claims (6)

1. The method of distortion of characters, including changing the graphic display of characters by changing the font used, characterized in that when it is implemented: (a) use at least two fonts containing images for at least the desired set of characters; (b) for each symbol, the procedure for creating intermediate images of a predetermined number of intermediate states is carried out by interpolating (morphing) each symbol image from one font into the corresponding symbol image from another font; (c) use the created intermediate images to graphically represent the characters. 2. The method according to claim 1, in which step (b) comprises the step of saving the resulting set of created intermediate images in the form of one or more new fonts, characterized in that the stored fonts are used in step (a). 3. The method according to claim 1, containing the use of at least one font, characterized in using at least one geometric shape to interpolate (morph) each character from the font to the geometric shape or vice versa.