Image Recognition for Solving Google’s reCAPTCHA : An Investigation of how Different Aspects Affects the Security of Google’s reCAPTCHA

Sammanfattning

In this project, an image recognition solver for Google’s reCAPTCHA was used to investigate how different aspects of the reCAPTCHA challenge affects its efficiency on defending against bots. The aspects that were examined was the type of reCAPTCHA challenge and the object which should be identified in the images. To conduct the study, an image recognition algorithm, trained on a large image dataset, was used in a script that interacted with a reCAPTCHA website. Testing showed that both the investigated aspects had a large impact on the efficiency, individually and in combination.