How to Bypass Captcha with a Machine?

How to Bypass Captcha with a Machine

The TT situation i.e. the Turing Test situation depends upon the interrogator in a remote chatting room with a machine and a human with some other network. The interrogator asks a few questions to both human and machines and checks the answers.  If the interrogator is able to correctly distinguish between human and machine by virtue that the human’s answer is more precise then the examined system is not intellectual. According to the May 2014 edition vol.10, No.15 ISSN: 1857-7431 193 of the European Scientific Journal Edition says that no artificial intelligence (AI) system is yet to pass the Turing Test by using a language as an indicator of intelligence. Language is one among the best indictors of intelligence. Apart from Language even pattern recognition is another good indicator for intelligence. In fact the questions in various IQ tests depend completely on the pattern recognition to assess human intelligence. Another sign of acumen is playing the game of chess in the master’s level. In this situation we can argue that any AI system can perform equivalent to a human or even better that a human is known as an intelligent system. This means that the majority of AI systems are planned to deal with issues that require extensive searching and a lot of computations which usually defeat the human capabilities since machines can perform calculations much faster than humans.  This can be used for a huge number of computer games but it cannot be used on games like chess, n-puzzle, cube puzzle, n-queens and crossword puzzles. Therefore it is believed that AI is a much more intelligent when compared to humans especially when it used on particular domain.

The main idea of Captcha is to differentiate between humans and machines in security applications like preventing posting automatic comments and registering on various websites, safeguarding e-mail ids from web scrapping and protecting online polls from being distorted by firm views. There are many types of Captcha’s like Gimpy and EZ Gimpy, which are based on OCR. An example of a simple image processing is shown below. A simple image processing gives nearly a perfect segmentation which actually leads to perfect recognition because segmentation is the main problem when dealing with such tasks.

Simple OCR System with TT

In this planned system a Captcha is created automatically with randomly selected character to create an image. The image is then seriously tarnished and sent through a network to the human and a client is ready with a simple OCR system. This image is read by both humans and computers and send the text back to the integrator. An example of Simple OCR system with TT is given below.

The decision of the interrogator depends upon the accuracy. The most precise answer is said to have come from the human. If the interrogator is unable to differentiate between the answers i.e. is the systems performance is equivalent to the human’s performance then the system is intelligent in that particular task. In the picture given below you can see the Process of Creating Captcha out if random characters XXYH.

The final Captcha is then sent to all the clients including the OCR machine through a local network. Both the humans and the machine read and break the received Captcha and then send the text to the interrogator who verifies the answers based upon their accurateness. In the picture below you can see the series of image processing and analysis performed by the OCR machine.

The machine is educated on English alphabets with the help of an artificial neural network. Each part is the input to the ANN to find the best node index which is linked with a particular character. All the character are then collected and sent back to the interrogator.

Mortifying images by adding noise and breaking character makes it tough but not impracticable to solve Captcha’s with OCR programs. Nevertheless, making it difficult for a computer to recognize also makes it tough for humans. The design of Captcha should not depend upon the reverse TT. On the other hand we argue that TT is a good intelligence test especially for machines that are involved in a particular job.

Leave a comment