When you hear about Artificial Intelligence, you might think of robots with human-like characteristics. That’s where AI science fiction meets reality. AI is a study that aims to create machines that can think, learn and act like humans. For example, computers are programmed to play chess and drive cars. These AI programs rely heavily on two essential tools: deep learning and NLP. Using these technologies, systems can be trained to achieve certain tasks by processing large amounts of data and recognizing patterns. Here is How AI can Guess Your Password.
Process of AI Guessing your Password
AI solutions are learning to guess your password and are probably very good at getting it. We think AI will soon crack complicated passwords more effectively than any human hacker. Machine learning techniques allow a computer algorithm to learn and improve from data it processes. If an algorithm is designed to match patterns in data, like recognizing photos or deciphering text, then feeding an algorithm with millions of examples can produce a system that rivals human intelligence. Using these techniques, machine learning algorithms tell us that humans are predictable.
We all know that using a simple password like “admin” is bad. But, conversely, using a more complex word like “correcthorsebatterystaple” is less risky. But what if you had to choose between them?
If you have a strong password, you can expect it to be secure until a data breach compromises your password at a site with weak security. Therefore, you should use a strong password to avoid data breaches through AI.
Researchers at SIT (Stevens Institute of Technology) in New Jersey and New York Institute of Technology have created a deep learning tool to guess passwords with surprising accuracy. As a result, they can guess passwords with incredible accuracy.
Guessing Tools
Compared to state-of-the-art, rules-based password guessing tools such as HashCat and John the Ripper, ‘PassGAN’ proved more effective at cracking passwords, according to test results released by Researchers.
The researchers matched nearly 47% of passwords from a testing set of actual user passwords leaked after a breach at RockYou in 2010. PassGAN performed better than John the Ripper in their evaluation, achieving a factor of two increased password cracking speed.
Combining output of PassGAN with HashCat, researchers could match 24% more passwords than they could by using either tool alone. In addition, researchers said their technique was faster and could churn out many more passwords than other methods. Paolo Gasti, one of the authors and a researcher, says their research shows an “exciting proof of concept.
GANs
GANs (generative adversarial networks) are neural networks that create new data similar or nearly identical to their source data. For example, GANs use authentic images to learn how to generate realistic images of people and animals.
A GAN comprises two deep neural networks that “interact continuously.” Gasti uses a method of an eyewitness and a sketch artist to illustrate how two networks function.
A sketch artist might start with a rough outline of a suspect, then refine an image using eyewitness information until they have a very accurate picture.
Gasti says that a discriminative network uses data to train a generative network to produce increasingly similar images to actual samples.
Gasti, researchers Briland Hitaj, Giuseppe Ateniese, and Fernando Perez-Cruz from Stevens Institute of Technology in Hoboken, New Jersey, said they wanted to see if a similar thing could be leaked passwords.
PassGEN was designed to see if a machine-learning tool could create passwords by looking at and learning from real people’s passwords.
The researchers fed what they suspected were leaked RockYou passwords into PassGAN to see if a program could generate new, realistic-looking passwords. The researchers then checked this result against rest of leaked dataset and found that 47% of data could be matched. So, in another way, 47% of passwords generated by PassGAN would have worked against RockYou accounts for which they were initially developed.
Gasti says that while other tools use structured rules and patterns to guess passwords, PassGAN is better at generating random combinations of letters, numbers, and symbols. In future, it will be engrossing to see how approaches like PassGAN will continue to improve with larger datasets and better computing power. As datasets and computing power continue to grow, we will see how systems like PassGAN improve.
Passwords can be guessed by machines trained to identify common patterns in future. Gasti says that such a development could force organizations that rely on passwords to use more sophisticated forms of authentication. However, he has mixed feelings about use of such tools. If we can figure out how to use these systems, someone else will likely be able to.
Conclusion
It’s essential to set up a good password. However, it is also important to remember your password. In some cases, two are in direct opposition to each other. The best way to combat that is to use a password manager (Lifehacker is one of our favorite sites for password suggestions) and passwords that aren’t related to personal details.
Email is a significant step toward making password management easier for end-users. Websites can test a user’s password’s strength by trying it against potential employees’ email accounts. A great addition to any company account management processes, this technology could significantly help prevent many hacking attempts in the immediate future, as support for two-factor authentication and fingerprint recognition increases across platforms in coming years.
As we see more AI applications, it’s essential to understand these systems’ security and vulnerabilities. AI is moving quicker than us and acting in ways we hadn’t anticipated, so it’s up to us as a human race to stay one step ahead. AI is here to stay as long as we continue down this road of innovation and curiosity, so let’s do our best to forge ahead into world of chatbots and voice assistants and AI Computer Vision Solutions without forgetting how they work and what they’re capable of.
About Author
Am Suresh working as a Senior Technical Content Writer at Visionify.ai and written research content on AI, ML, and computer vision for Visionify. Develops computer vision solutions for many food manufacturing and retail. Trends in Artificial Intelligence make me dive into it like #artificialintelligence on LinkedIn, where I get inspired by tones of data on AI.
