The study has identified a novel method of cyber attack that involves eavesdropping on keyboard sounds to steal personal data. Researchers utilized a homemade deep learning-based algorithm to analyze these sounds and successfully decoded the input content. This attack exploits acoustic side channels, obtaining sensitive information by analyzing keyboard sounds. Researchers point out that hackers can easily obtain recordings through smartphone microphones or applications like Zoom, and decode them using the algorithm. The article discusses the threats posed by this attack and potential defense measures. This research is of significant importance for users to understand and protect their data security.