As we know, there are now more and more cyberattacks that have a serious impact on individuals, businesses, and governments. Attacks are increasingly sophisticated and modern, requiring competent entities to study and devise solutions to quickly detect and minimize user risks. In fact, attacks come in many different ways and many different environments. In this paper, we do an empirical study on the deep learning model using some kinds of activation functions applied for detecting attacks in the wireless local area network (WLAN) data. We also point out some discussions about activation functions for future research.