A team of researchers from several British universities has developed a self-learning AI that listens to keystrokes and can predict which words, including passwords, are being typed on a keyboard using a deep learning model.
Artificial Intelligence is now an integral part of everyday work and is taking on more and more tasks from us humans.
The prominent example ChatGPT is a chatbot that communicates with users through message exchanges. Through deep learning (a method of machine learning), the collected data can be used to generate increasingly better responses. The British research team follows the same approach with their AI, only in an acoustic manner.
The AI listens well and learns from it
To feed the technology with the necessary data, the research team used the keyboard of a modern MacBook Pro. Each of the 36 keys was pressed 25 times, and the resulting sounds were recorded.
Using an algorithm, the recordings were further processed, and the identifiable differences for each key were visualized as waveforms. With the data obtained this way, the researchers can now predict keystrokes with alarming certainty.
Only the recording method produces better or worse results.
A new digital danger
Any regular microphone can serve as a recording source. Even through phone calls or a smartphone with an installed meeting app, the AI can determine which keys are being pressed with up to 95% accuracy. This means that in the future, you not only have to be careful about what you say when the microphone is on, but also what you type.
You can protect yourself in this case with biometric authentication methods (e.g., fingerprint) or by using password managers.
What do you think about this development? Are you already actively using AI in your daily life, or can you not find anything positive about it?
Source: pcgamer.com
