The “ChatGPT of passwords” can crack any combination in less than a minute

Share This Post

- Advertisement -

THE Passwords they are the number one security factor on the internet, but people often struggle make them safe and easy to remember. In fact, studies ensure that people generally store up to five e’s in their memory then recycle the combinations across multiple accounts.

- Advertisement -

This common practice among most users leads to the use of weak passwords that cybercriminals can breach without too many complications, and thus keep private data, bank keys and more.

In this regard and as a security tool, researchers have developed PassGANa model of deep learning (machine learning) which helps to generate passwords and, at the same time, verify that they are really secure.

- Advertisement -

How PassGAN works

PassGAN, the artificial intelligence that helps generate secure passwords.

PassGAN, the artificial intelligence that helps generate secure passwords.

This website is actually a Generative Adversarial Network (GAN), the same technology behind ChatGPT, which uses a training dataset to learn patterns and generate passwords. Consists of two neural networks: a generator and a discriminator. The first creates new passwords, while the other Evaluate if a password is real or fake.

To train PassGAN, a real password dataset. However, using these presents a security risk. Therefore, the researchers used a publicly available data set on password leaks called RockYouwhich contains over 32 million passwords leaked in a 2009 data breach.

The researchers preprocessed the data by removing duplicates, commonly used passwords, and passwords shorter than eight characters. Also added synthetic passwords to increase key diversity.

For PassGAN, a password including symbols, uppercase letters and numbers can be cracked after 2 million years of testing.  Photo: capture.

For PassGAN, a password including symbols, uppercase letters and numbers can be cracked after 2 million years of testing. Photo: capture.

After preprocessing, the data set was split into training and testing sets. The first performed the function of training the PassGAN model, while the other was used to evaluate the model’s performance.

Looking for invulnerable passwords

During training, the generator attempts to create passwords similar to the ones in the training dataset, while the discriminator evaluates the generator’s output and provides information on how to improve the generator’s performance. This process continues as long as the system can generate indistinguishable passwords of effective passwords.

Researchers evaluated the performance of PassGAN and came to the conclusion that it helped generate much more secure passwords compared to those of the test data set.

All you have to do is create combinations of 11 characters or more that include uppercase and lowercase letters, numbers and symbols. This allows you to crack keys that PassGAN will take 365 years to crack. Whoever puts 18 characters will make that number 22 million years.

Despite its performance, the passwords generated by PassGAN may only be suitable for some applications or services. For example, some require users to create easy-to-type passwordswhich may not be accurate for the AI.

Source: Clarin

- Advertisement -

Related Posts