Joshua Harrison, Ehsan Toreni, and Maryam Mehrnezhad of Cornell University have recently developed an Artificial Intelligence (AI) capable of cracking passwords by simply picking up the sound of computer keyboards.
Meticulously trained to perceive the different resonances caused by each key being pressed by capturing its various sounds, the AI has a remarkably high accuracy of 95%. However, by employing this technology through Zoom, the algorithm can achieve a no less remarkable 93% accuracy.
AI That Requires Training to Decipher Different Types of Keyboards
Regarding this development, the online publication Tom’s Hardware has pointed out that this new technology is not yet applicable to all keyboards, since it needs to be trained specifically for each model.
In this regard, the AI needs a particular keyboard with a precise reference of which character is related to each key pressed. Training can be carried out, as mentioned before, via Zoom and with the help of a good microphone.
The Questions Raised by this New Technology
While the size of this achievement is undeniable, it raises multiple questions related to the privacy and security of each individual. This is key since its remarkable ability to decode sounds may well be used negatively and maliciously.
The risks of access to bank accounts, personal data, credit cards, and other sensitive information are aspects that many have already put on the scales. Issues that have already prompted recommendations from experts to switch to much stealthier keyboards.
This story was written in Spanish by José Lomelí in Cultura Colectiva News
