Artificial Intelligence is disrupting many domains. The latest ones are bot detectors and ad-blockers. Here we discuss how AI is being used to eliminate the need of Captcha, and to develop advanced ad blockers.
Captcha has been around for more than 15 years. It has been widely used online to tell computers and humans apart. We have seen different kinds of Captcha (obscured images, text, and the infamous “I am not a robot”) being used in online payment portals, email subscribers, and many other areas. However, it has received a lot of criticism, especially from disabled people. Some people criticized it for slowing down their routine tasks as well. According to Wikipedia, it takes the average person approximately 10 seconds to solve a typical Captcha. Furthermore, many circumvention techniques have also been developed.
Keeping frustrations caused by Captcha in mind, Google has now introduced a smart and invisible Captcha that uses AI to do its job. Based on machine learning techniques, this invisible Captcha, now a part of “advanced risk analysis engine,” will monitor users’ behavior on the web page and divert only suspicious traffic towards native puzzle-solving pages. While Google has not revealed its code, it appears that it uses a combination of deep learning and supervised/unsupervised learning to detect bots. It might have a feature set of how humans browse the web that it would use to match patterns of ongoing browsing activity. If features do not match, it would raise a red flag diverting them to native Captcha pages.
Ad blockers saved us from annoying ads on web pages, especially those pop ups. Despite resistance shown by publishers, ad blockers have continued to exist and are now legal. They use a list of features to detect ads on a web page. This list includes markup codes, scripts, common URLs and more. When we open a web page, ad blocker scans it and if these features are found, that component of the web page is blocked.
Advertisers and publishers are now getting smart and using sophisticated mechanisms to embed ads into a web page in such a way that ad blockers are unable to detect them. For example, ad blockers cannot block Web Sockets so YouTube ads are not blocked. Moreover, Facebook ads are not blocked because they have similar code structure as normal content.
Researchers at Princeton and Stanford have come up with a new ad-detection mechanism that uses computer vision and AI to detect ads. The idea is to teach machines how humans look at and perceive ads. This new blocker called perceptual ad blocker uses OCR (Optical Character Recognition) along with container searches to detect ads. FTC’s regulation, which forces publishers to label ads so that people can recognize them, makes the job of this blocker easier because if humans can recognize ads, so can this intelligent blocker. While the researchers have only used this mechanism to detect ads and not block them (to avoid showing bias towards either publishers or blockers), this perceptual ad blocker can be the best one to date.
These two examples show that AI is not just disrupting big domains. It is not taking over our jobs. It is also making our lives easier by making our already-developed techs even better.
About the Author
Mazhar Naqvi is a CS grad student with research interests in computer networks and security. He can be reached at firstname.lastname@example.org and you can follow him on LinkedIn at https://www.linkedin.com/in/mazharnaqvi