In a new study, researchers from the University of Surrey in the UK have trained a deep learning model that can steal data from keyboard keystrokes recorded using a microphone. The model, called CoAtNet, can achieve an accuracy of 95% in identifying the keys that are pressed, even when the audio is recorded from a distance of 17 centimeters.

To train CoAtNet, the researchers recorded the sounds produced by pressing 36 keys on a MacBook Pro 25 times each. The audio was captured using an iPhone 13 mini that was placed 17 centimeters away from the laptop. The researchers then used the audio data to train CoAtNet to identify the keys that were pressed.

CoAtNet works by first converting the audio data into a spectrogram, which is a visual representation of the sound. The spectrogram is then used to train a convolutional neural network (CNN) to identify the keys that are pressed. The CNN is able to learn the unique sound signature of each key, and it can use this signature to identify the key that is pressed, even when the audio is noisy or distorted.

The researchers tested CoAtNet on a variety of datasets, and they found that it was able to achieve an accuracy of 95% in identifying the keys that were pressed. The model was also able to work with different types of audio, including audio that was recorded from a distance of 17 centimeters and audio that was recorded through a video conferencing software such as Zoom or Skype.

The researchers say that CoAtNet could be used to steal sensitive information like passwords and credit card numbers by simply listening to what a person is typing. They also say that CoAtNet could be used to spy on people in video conferencing calls.

The researchers have published their findings in the journalIEEE Transactions on Information Forensics and Security. They say that they are working on ways to improve CoAtNet's accuracy and to make it more difficult to detect.

In the meantime, users should be aware of the risks of having their keystrokes recorded. They should use a keyboard cover or a privacy screen to protect their typing from prying eyes. They should also be careful about what they type in public, especially if they are using a shared computer.

Here are some tips to protect yourself from this type of attack:

• Use a keyboard cover or a privacy screen to protect your typing from prying eyes.
• Be careful about what you type in public, especially if you are using a shared computer.
• Use a strong password manager to generate and store your passwords.
• Keep your operating system and software up to date with the latest security patches.
• Be aware of the risks of using public Wi-Fi networks.
• Use a VPN when connecting to public Wi-Fi networks.

By following these tips, you can help to protect yourself from this type of attack.