Researchers at Cornell have identified a way to steal data by recording keystroke audio. According to researchers, they have found that AI tools can be effectively used to compete in audio side-channel attacks during video calls.
New research has proven that AI can detect passwords and other forms of data by listening to the sounds that different keys make when pressed. Moreover, it can also guess the victim’s password during video conferencing calls.
The situation is very alarming, especially in light of the number of apps an average user allows microphone access to. Not only this, but the extensive use of voice-over apps like Zoom makes it easier for AI to discover passwords and other sensitive data.
It is one of the most recent kinds of AI scams, frauds that are directly based on deep-learning models that can recognize keystrokes in this way.
AI Can Easily Identify The Keys You Press On Your Keyboard
A recent study conducted by UK-based researchers Joshua Harrison Ehsan Toreini and Maryam Mehmezhad found that a deep learning model can identify keystrokes with 95% accuracy made on a 2021 MacBook Pro recorded on a nearby iPhone 17 mini.
To train the model to identify the individual waveforms produced by individual KeyStrokes, they checked out by pressing 36 keys 25 times each.
According to the researchers, it showed impressive and outstanding results by the iPhone audio recording “when trained on keystrokes recorded using the video-conferencing software Zoom, an accuracy of 93% was achieved, a new best for the medium”.
Surprisingly, the deep learning model used by the researchers was mistaken in its classification and was just a few keys away from the correct answer. Thai shows that the position of the keys on the keyboard plays a vital role and affects the creation of distinctive audio profiles.
As per the researchers, “prove the practicality of these side-channel attacks via off-the-shelf equipment and algorithms.”
Some Side Channel Attacks
Side-channel attacks are a kind of cyber attack that takes advantage of extra information that is present through different mediums and channels created by the physical implementation of the system and its hardware.
One instance of a side-channel attack is the capturing of audio in a video call. Van Eck Phreaking is another kind of side-channel attack, that involves utilizing equipment that can detect electromagnetic emissions emitted by LCDs to extract data to know what is on the screen.
How To Protect Yourself From The Threat
There are some ways by adopting them you can prevent yourself from threats and attacks facilitated by AI tools. There are some suggestions with the abovementioned capabilities to combat the cyber attacks facilitated by AI tools suggested by three researchers.
The most effective method to implement actions is to simply use multiple cases in your password. The AI tool is good at recognizing most keystrokes when typing while on video calls. You can even use a password manager to help juggle various combinations are highly recommended by the trio of researchers. It is one of the ways to keep yourself safe from threats and limit the damage if one of your accounts is compromised.
Read more:
AI Technology: Tech Companies Recession Buster
YouTube Starts Testing AI Based Video Summarizer
