Here is my project of research On Keyboard Acoustic Eavesdropping and Typing Speed in Skype. Here is the abstract of this thesis: In this paper, we utilize a small keyboard sound dataset to classify key sounds information using traditional machine learning method. The effects of typing speed and style on model accuracy are explored by simulating call scenarios in Voice-over-IP software, Skype. On the test set, our results show that our model performs well in terms of accuracy, achieving 86.73% accuracy and 97.62% in terms of Top-5 accuracy. At the same time, we delved into the effects of typing speed and typing style on accuracy and found that an increase in typing speed impacts model accuracy. Finally, we propose two possible methods to defend against this voice attack style to help protect personal privacy. These methods aim to enhance cyber security and reduce the risk of potential privacy breaches.
in the future, i will add the autoencoder to denoise when we face the situations in various noise scenarios.