Field of the Invention

The invention is related to a "Completely Automated Public Turing test to Tell Computer and Human Apart" (throughout the following abbreviated by "CAPTCHA"), and in particular to a method for providing a CAPTCHA employing dynamic images and a system therefor.

Background of the Invention

Nowadays, the life without the Internet and online services is not imaginable. Internet sites and online services are growing daily, and respectively, the related dangers are growing in parallel. One of these kinds of dangers is kind of Internet bots, which try to pose themselves as a human and act like a human. An Internet bot (in the following also shortly termed "bot") can perform simple and structured tasks, which it is programmed for, in many times faster than a human. It can, for instance, create a free e-mail account (or many e-mail accounts) for a likely malicious (or even commercial) purpose. Bots can also disturb an online appointment system or voting system by filling forms thousands of times to a certain option. To cope with such a problem, many Internet service providers (e. g. Google) use a technology named CAPTCHA (Completely Automated Public Turing test to Tell Computer and Human Apart). The CAPTCHA technology employs programs that protect web based services against malicious bots by generating and grading tests that humans can easily pass, but current computer programs cannot or increase their costs [4]. For instance, a human can read distorted text (as illustrated in Figure 1), but this simple action would be difficult for current programs to solve. More information thereon can be found in the patents listed by Refs. [1 - 3, 5]).

The CAPTCHAs are used in several web based services in different stages like creating accounts, reseting passwords or against dictionary attacks to any online account during login process. In addition, it is used in wide varity of online sevices like poles, appointmnets, news groups, blogs, etc.

State of the art

The following list itemizes various state-of-the-art types of CAPTCHAs:

• Text based: this type can be realized by several implementation models. Figure 1 illustrates a typical implementation used by Google.
• Graphic based (e. g. Bongo project [15]): Displays two series of blocks, wherein the user must then find the characteristics that sets the two series apart into two different categories e. g. all element in the right is drawn with thick lines. The user is asked to determine, which series - each of four single blocks - belongs to which category.
• Graphic based (e. g. Pix project [16, 17]): In this method, a large database of labeled images is used. Each time an object is randomly selected and four images related to that object are randomly picked from the images database, and then the images are distorted. Users will be asked to pick the object which is the subject of images form a list of words.
• Audio based: In this CAPTCHA, a word or a sequence of numbers is picked randomly from a database and rendered in an audio clip using a TTS software. Then, the speech signal files is degraded by common degradations (like codecs, packet loss, and background noise) and presented to the user. The user is asked to identify and type the word or numbers he has heard.
• Game based: There are many implementation types of game based CAPTCHAs, e. g. puzzle CAPTCHA. Usually, in a puzzle CAPTCHA a given picture is divided to chunks. A user is supported to combine these chunks so as to form the complete picture, same as the original one.
• Video based: It is a newer and less commonly seen CAPTCHA system. In video-based CAPTCHAs, three words (tags) are provided to the user which describes a video. The user's tag must match to a set of automatically generated "ground truth tags": only then, the test is said to be passed. The term "video CAPTCHA" is used to any CAPTCHA that uses a video in order to present information to a user.
• Advertise based CAPTCHA: The goal of this method is to create a fun interactive technology, which delivers a fully captured audience for advertisers to brand their message to without compromising the security of the publisher's website. In other words, this approach gives the opportunity to the advertisers to indicate (advertise) their product (or message, motto, etc.) without breaking the security rules of publisher's website. As an incentive to publishers, ADSCAPTCHATM will share advertising revenue with its publishers, so they can continue stop spam while making money [12].

Note that in case that any degradation on the content is used (on image or audio video), the level of difficulty of test is increased by increasing the degradation parameters (like amount of noise, radiation, etc.). The CAPTCHA service providers use those parameters to increase the difficulty in case that they suspect that the requested user is a potential bot. It leads to low quality of experience for real users.

A further widely used CAPTCHA type known from the prior art is the concept of "reCAPTCHA". Detailed information thereon can be found in Refs. [6] and [11]. Google describes the concept as follows [10]:

• "Millions of CAPTCHAs are solved by people every day. reCAPTCHA makes positive use of this human effort by channeling the time spent solving CAPTCHAs into digitizing text, annotating images, and building machine learning datasets. This in turn helps preserve books, improve maps, and solve hard AI problems."

Further according to Google [10], main applications of the reCAPTCHA concept are:

• "reCAPTCHA helps solve hard problems in Artificial Intelligence. High quality human labelled images are compiled into datasets that can be used to train Machine Learning systems. Research communities benefit from such efforts that help build the next generation of groundbreaking Artificial Intelligence solutions."
• "reCAPTCHA digitizes books by turning words that cannot be read by computers into CAPTCHAs for people to solve. Word by word, a book is digitized and preserved online for people to find and read."
• "reCAPTCHA improves our knowledge of the physical world by creating CAPTCHAs out of text visible on Street View imagery. As people verify the text in these CAPTCHAs, this information is used to make Google Maps more precise and complete. So if you're a Google Maps user, your experience (and everyone else's) will be even better."

Whereas the above described types of CAPTCHAs proof powerful in resisting fully machine driven attempts to pass the CAPTCHA, they may fail when humans at least partly contribute to these attempts. One possibility to systematically deploy human intelligence in order to pass CAPTCHAs makes use of "crowdsourcing". The crowdsourcing micro-task platforms provide human intelligence as a service through API. This will be explained in the following.

Attacking CAPTCHA/reCAPTCHA via crowdsourcing services

According to Ref. [13], "crowdsourcing is an innovative way of organizing jobs and workers. The collaborative brainstorming enables hundreds or even thousands of people to contribute their thoughts and energies to a single task and can support complicated jobs that conventional means could never manage. Crowdsourcing uses an Internet task market, usually called a crowdsourcing platform, to connect workers to jobs. It enables those workers to take the kinds of jobs that they like and to complete those jobs when and where they want. Employers can find the kinds of workers they need, with the specific skills for their jobs, and pay for only the amount of work they need."

Thus, it is possible to include human intelligence, using a crowdsourcing micro-task platform API, in the attack process such that the bot tries to fraudulently or maliciously make an image from the CAPTCHA/reCAPTCHA question and automatically deliver it to an unaware "human", who is working on a normal task in crowdsourcing platform, and use his/her answer to solve the CAPTCHA/reCAPTCHA question. Then, one will be faced with a "human/machine" cooperation, which can break the security that has been putting on under attacked web base services. In particular, the questioner model of reCAPTCHA could also be learned by the techniques of machine learning and may eventually be circumvented.

As a consequence, there is a need for the provision of a CAPTCHA that cannot be circumvented (i. e., reliably be passed) by techniques employing crowdsourcing, or that at least considerably impede attempts to pass the CAPTCHA by those techniques.

The object of the present invention is thus the provision of a method and a system that provide a CAPTCHA that cannot be circumvented (i. e., reliably be passed) by techniques employing crowdsourcing, or that at least considerably impede attempts to pass the CAPTCHA by techniques using crowdsourcing.

This object is achieved by the method and the system with the features as disclosed by the claims presented in this document.

Summary of the Innovation

The present invention - termed "Dynamic Image CAPTCHA" (in the following also shortly referred to as "DIC") - employs techniques (e. g. dynamic images, time limitation as a technique for cope with the threats of crowdsourcing) to tackle the above-identified problem with any kinds of the Internet bots. In the following, a new method for a more secure CAPTCHA system based on dynamic imaging will be described, which comprises countermeasures also against social media hacking.

One aspect of the invention relates to a method for providing a CAPTCHA (completely automated public Turing test to tell computers and humans apart). The method comprises the steps:

• (1a) sending, by a browser or web application of a client, a request for providing a CAPTCHA to a server;
• (1b) generating, on a server, a set of pictures, the set containing at least two pictures, wherein each of the pictures is assigned a unique identifier, and a temporal identifier which is generated by each request based on the unique identifier; and sending the set of pictures and the temporal identifiers, and the assignment between temporal identifiers and pictures to the browser or web application;
• (1b') generating, on the server, a question that is associated with one or more of the pictures of said set, and wherein the question is assigned a question identifier; sending the question and the question identifier to the browser or web application; and storing, on the server, the temporal identifiers assigned to the pictures, with which the question is associated, and the question identifier;
• (1b") distributing the pictures of the set to one or more subset(s) of pictures, wherein each of the subsets comprises at least two pictures;
• (1b"') defining, for each of the subsets, a presentation position within the browser or web application, and generating, for each of the subsets, an enumeration for the pictures of the respective subset;
• (1c) presenting, in the browser or web application, the question;
• (1d) presenting, consecutively, in the browser or web application, for each of the subsets at the position defined in step (1b"') any of its pictures for a predetermined presentation time, wherein the pictures of the respective subset are presented in the order of the enumeration of the pictures in that subset,
• (1e) wherein, for each of the pictures or subsets presented, the browser or web application detects whether or not the user performs a confirmation action during the presentation time of the picture, and
• (1f) wherein, when the browser or web application has detected a confirmation action, the temporal identifier(s) assigned to the presented picture(s) while the confirmation action has been performed and the question identifier are sent to the server;
• (1g) if one or more temporal identifier(s) and the question identifier have been received by the server from the client: checking, by the server, for identity of the temporal identifier(s) sent to the server in the foregoing step and the temporal identifier(s) stored in step (1b') for the question identifier received in this step;
• (1h) if identity has been confirmed in the foregoing step: sending, by the server, information that the CAPTCHA has been successfully passed to the to the client; otherwise: sending, by the server, information that the CAPTCHA has not been successfully passed to the client.

Here, a "set of pictures" (as generated in step (lb)) denotes an unordered collection of pictures. An order of the pictures of such a set (or subset of the set) is then established in step (1b"') by the generation of an "enumeration". However, as the enumeration is not intrinsic to the set, the pictures of a set can be re-enumerated when this appears appropriate.

The term "presentation time" refers to the time a certain picture is shown to the user, when the pictures of the set of pictures are consecutively presented to the user.

The expression "confirmation action" relates to any action of a user that is performed during the presentation time of a picture, and which is suitable to notify the browser or web application that the shown picture shall be chosen by the user. Preferably, the user chooses a picture, when he/she assumes that the picture represent the "true" answer on the question or request he/she has been asked before (in step (1c)).

In one embodiment, the method is performed such that in step (1b), the set of pictures is generated by choosing a set from a pool of fixed sets of pictures being held available on the server.

In one embodiment, the method is performed such that in step (1b), the set of pictures is generated by choosing at least two pictures from a pool of pictures being held available on the server.

In a preferred embodiment of the method, the method is performed such that:

• in step (1b"), the distribution of the pictures to the subsets is carried out randomly; and/or
• in step (1b"'), the definition of the presentation positions for the pictures of the subsets is carried out randomly; and/or
• in step (1b"'), for any one of the subsets, the enumeration for the pictures is generated in a random fashion.

Preferably, in step (1b"), the distribution of the pictures to the subsets is carried out by the server, and step (1b") comprises the further substep of:

• sending, by the server, the arrangement of the pictures defined by the subsets to the browser of web application; and/or

wherein in step (1b"'), for at least part of the subsets, the enumeration for the pictures is generated by the server, and wherein step (1b"') comprises a further substep of:

• sending, by the server, the generated enumeration(s) of the pictures to the browser or web application.

Further, step (1b") and/or step (1b'") may be performed by the browser or web application.

Also, when all pictures have been presented without detection of a confirmation action in step (1e), steps (1f) to (1h) may be replaced by the following step:

• finishing the CAPTCHA and handing over to the browser or web application the information that the CAPTCHA has not been successfully passed.

Note that in the latter case, steps (1f) to (1h) may be without any effect.

Moreover, when all pictures have been presented without detection of a confirmation action in step (1e), steps (1d) and (1e) may be repeated,

• wherein the number of repetitions is preferably limited, more preferably limited to a maximum number of 2 repetitions, and most preferably limited to a maximum number of 1 repetition.

According to one embodiment, for each of the subsets, the number of pictures in the subset is between 2 and 20, preferably between 3 and 10, and most preferably 4 or 5.

According to one embodiment, the presentation time for each of the presented pictures is between 0.5 to 10 seconds, preferably between 1 to 5 seconds, and most preferably between 2 to 4 seconds.

When a predetermined maximum time is exceeded, the following step may be performed:

• finishing the CAPTCHA and handing over to the browser or web application the information that the CAPTCHA has not been successfully passed.

In one embodiment of the method, the followings steps may be performed:

1. (i) repeating step (1b) with the additional condition that the question or request is associated with at least one of the pictures of the newly generated set of pictures,
   
   

   wherein the new set of pictures is different from the previously generated set(s) of pictures;

2. (ii) repeating the steps of the method beginning from step (1b"),

wherein the repetition according to any one of steps (i) and (ii) is limited to a predetermined maximum number, the maximum number being set preferably to 1, 2, or 3.

In one embodiment of the method, the following step may be performed:

1. (a) repeating the steps of the method beginning from step (1b'),

wherein the new question generated in step (1b') is different from the previously generated question(s),

wherein the repetition according to step (a) is limited to a predetermined maximum number, the maximum number being set preferably to 1, 2, or 3.

In one embodiment of the method, the following step may be performed:

1. (A) repeating the steps of the method beginning from step (1b),

wherein the new set of pictures is different from the previously generated set(s) of pictures,

wherein the new question is different from the previously generated question(s), and

wherein the repetition according to step (A) is limited to a predetermined maximum number, the maximum number being set preferably to 1, 2, or 3.

In a further embodiment of the method, step (1f) comprises additionally: sending a secret code of the client to the server; and step (1 g) comprises additionally: receiving the secret code and using the secret code for authenticating the client.

Here, a "secret code" is an identifier of the client which is kept secret on the client side machines and bound to them. It is used for authenticating the client.

A further aspect of the invention relates to a system for providing a completely automated public Turing test to tell computers and humans apart (CAPTCHA), according to the method described above. Said system comprises:

• a server; and
• a client, on which a browser or a web application is installed.

In particular, the server is configured for:

• generating a set of pictures, the set containing at least two pictures, wherein each of the pictures is assigned a unique identifier, and a temporal identifier which is generated by each request based on the unique identifier;
• sending the set of pictures and the temporal identifiers, and the assignment between temporal identifiers and pictures to the browser or web application;
• generating a question that is associated with one or more of the pictures of said set, and wherein the question is assigned a question identifier;
• sending the question and the question identifier to the browser or web application;
• storing the temporal identifiers assigned to the pictures, with which the question is associated, and the question identifier;
• if one or more temporal identifier(s) and the question identifier have been received by the server from the client: checking for identity of the temporal identifier(s) sent to the server and the stored temporal identifier(s);
• if identity has been confirmed: sending information that the CAPTCHA has been successfully passed to the to the client;
• otherwise: sending, by the server, information that the CAPTCHA has not been successfully passed to the client.

Further, the browser or web application is configured for:

• sending a request for providing a CAPTCHA to a server;
• presenting the question;
• presenting, consecutively, for each of the subsets at a predefined position any of its pictures for a predetermined presentation time, wherein the pictures of the respective subset are presented in the order of a predefined enumeration of the pictures in that subset,
• wherein the browser or web application is configured for detecting, for each of the pictures or subsets presented, whether or not the user performs a confirmation action during the presentation time of the picture, and
• wherein the browser or web application is configured for sending to the server the temporal identifier(s) assigned to the presented picture(s) while the confirmation action has been performed and the question identifier, when the browser or web application has detected a confirmation action.

In one embodiment of the system, either the server or the browser or web application is configured for:

• distributing the pictures of the set to one or more subset(s) of pictures, wherein each of the subsets comprises at least two pictures.

In one embodiment of the system, either the server or the browser or web application is configured for:

• defining, for each of the subsets, a presentation position within the browser or web application.

In one embodiment of the system, either the server or the browser or web application is configured for:

• generating, for each of the subsets, an enumeration for the pictures of the respective subset.

Other aspects, features, and advantages will be apparent from the summary above, as well as from the description that follows, including the figures and the claims.

List of figures:

Fig. 1:

Example of a typical CAPTCHA known from prior art;

Fig. 2:

Example of adsCAPTCHAs;

Fig. 3:

Graphic CAPTCHA;

Fig. 4:

Audio CAPTCHA;

Fig. 5:

Game CAPTCHA;

Fig. 6:

Example of "reCAPTCHA";

Fig. 7:

Example of "Dynamic Image CAPTCHA" according to the invention.

Dynamic Image CAPTCHA (DIC)

In the following, it will be described how to cope, according to the invention, with the above-identified problems. In the system and the method of the invention (DIC technology), the DIC service provider has a server with many dynamic pictures; these pictures can have any content (commercial, pedagogical or whatever a provider wants to show). Depending on the topics of the images, there will be a question and a single answer, but the user will see a set of N pictures, which is changing periodically (meaning every x seconds, another set of images will be shown). The user will be asked, for instance, to click on the "true" image(s) in the image set, i. e., on one (or more) of said N pictures that have been pre-assigned to the question. In other words, as the picture set is shown in chronological order, the user has to wait till he/she sees the true picture and then click on it. The periodical changes of image set leads to tackle any abuse with crowdsourcing technology. As time limitation on DIC expires, answers will not be accepted anymore and:

1. 1) A new set of pictures belonging to the same question would be loaded and shown to the user. Of course, the order of pictures is random and is changed after each load. For example, if the true option for a question was the second picture among all, after passing the time dead-line and load a new set of pictures, the correct answer may place as the forth, third or even again the second picture.
2. 2) A different question corresponding to the shown pictures would be loaded, i. e., the set of pictures does not change, but the correct answer (picture) does.
3. 3) A complete new question with a new set of pictures would be taken and shown to the user. Here, the same above mentioned rule would be used, i. e., the order of pictures (regardless of the question and/or the set of pictures) would always dynamically/randomly change.

The combination of a DIC as above and time limitation makes it possible to provide a new human-computer detection test, which can be both, safer against bots and more attractive for any online service provider (or even business companies). The idea is that one needs not to be worried, if the HTML code of this CAPTCHA would be available for everyone, because the important point here is the approach being employed to secure the CAPTCHA. This approach, an example of the DIC technique according to the invention, is preferably based on a main DIC server, which holds a number of diverse dynamic pictures (and corresponding questions). In the DIC server, each picture has a unique identifier, according to which the server later can decide, whether or not the image chosen by the user is the correct one. More precisely, the server knows that the image with a certain unique identifier (e. g. TE1236Xr1) is the correct answer. Note that each time a new temporary identifier is created for each image before sending to the client. Therefore, a client does not reveal the real identifier of specific image. When the user clicks on one of the images, the corresponding (intermediate) code/identifier is sent back to the DIC server, and then the server compares the returned identifier with that one it knows as the correct answer. If they are equal, the task is successfully done, and if not, a new set of pictures or question depending on the implementation policy would be taken.

An example of the DIC approach according to the invention is shown in Fig. 7. In the table of Fig. 7, four different sets of pictures A, B, C, and D are shown, each of which comprising four stickers (emoticons) depicted in a column of the table. Of course, all shown pictures should be changed dynamically. As in a printed document, the motion of pictures cannot be simulated, the pictures are shown at 4 different time intervals T1, T2, T3, and T4 being represented by the rows of the table of Fig. 7. The time intervals T1 to T4 may preferably be of equal time length, e. g., 1 second. For any of the sets of pictures A, B, C, and D, only one picture is assigned uniquely to a question that has been created for the respective set of pictures, this picture being termed the "true picture" (of the respective set). Accordingly, only in one of the shown pictures the user may recognize the true picture and he/she has 1 second to click on it.

As to the first set of pictures (i. e., the set shown in column A), a likely question could be: "Question 1 - please select all pictures containing smiles!" Then, the user should click within the third second (T3) after the DIC has been started (marked by the arrow in the field given by column A and row T3).

The (funny) question associated with the second set of pictures (set depicted in column B) may be: "Question 2 - which picture represents a person who has children in teen age?" The answer is then given by the "true picture" depicted at column B/row T2 (again marked by an arrow).

While the invention has been illustrated and described in detail in the drawings and foregoing description, such illustration and description are to be considered illustrative or exemplary and not restrictive. It will be understood that changes and modifications may be made by those of ordinary skill within the scope of the following claims. In particular, the present invention covers further embodiments with any combination of features from different embodiments described above and below.

Furthermore, in the claims the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single unit may fulfil the functions of several features recited in the claims. The terms "essentially", "about", "approximately" and the like in connection with an attribute or a value particularly also define exactly the attribute or exactly the value, respectively. Any reference signs in the claims should not be construed as limiting the scope.

References

1. [1] http://www.google.de/patents/DE20201310091 OU 1 ?cl=en
2. [2] Wang, James Z., Ritendra Datta, and Jia Li. "Image-based CAPTCHA generation system." U.S. Patent No. 7,929,805. 19 Apr. 2011.
3. [3] http://www.google.com/patents/US8483518
4. [4] http://www.ijcsit.com/docs/Volume%205/vol5issue02/ijcsit20140502289.pdf
5. [5] https://www.google.com/patents/LTS20130091027?dq=advertising+captcha&hl=de&s a=X&sqi=2&pjf=l&ved=OahUKEwio60jr2bDJAhUMEiwKHeJkD_8Q6AEIHDAA
6. [6] http://www.cs.cmu.edu/∼biglou/reCAPTCHA_Science.pdf
7. [7] http://www.captcha.net/captcha_crypt.pdf
8. [8] http://www.nytimes.com/2002/12/10/science/human-or-computer-take-this-test.html
9. [9] http://radar.oreilly.com/2007/05/recaptcha-stop-spam-read-books.html http://www.wired.com/2007/06/ff-humancomp/
10. [10] https://www.google.com/recaptcha/intro/index.html
11. [11] http://www.google.com/patents/US20120023549
12. [12] http://ww1.prweb.com/prfiles/2010/10/27/4176174/NewadvertisingtoolbasedonCAP TCHA.pdf
13. [13] http://www.dummies.com/how-to/content/crowdsourcing-for-dummies-cheat-sheet-uk-edition.html
14. [14] http://www.dummies.com/how-to/content/crowdsourcing-advantages-five-things-crowds-do-wel.html
15. [15] http://www.captcha.net/captcha_cacm.pdf
16. [16] http://www.ijsce.org/attachments/File/v2i5/E1038102512.pdf
17. [17] http://www.captcha.nedcaptchas/pix/