Detection of LINK 16 signal, i.e. identifying the presence/absence of this system in battlefield, is of great interest. In this paper, a new method for detection of LINK 16 signal is proposed with a very lower complexity and better performance compared to the current methods. Due to the periodic nature of transmitted energy, three different features are defined and extracted which result in a detection probability more than 80% for the values of signal to noise ratio around -16 dB. Since this method extracts suitable features from the sampled signal and applies simple rules which are obtained via machine learning, it can be realized in real-time practical detection systems.