Abstract. The wide deployment of 802.11 WLANs has led to the coexistence of wired and wireless clients in a network environment. This paper presents a robust technique to detect 802.11 wireless hosts through passive observation of client traffic streams at the edge of the network. It is based on the estimation of entropy of packet interarrival times and on the analysis of variation in the measured entropy values across individual end host connections. With the aim of generating a physical layer “signature” that can be easily extracted from packet traces, we first perform controlled experiments and analyse them through Spectral Analysis and Entropy evaluation. Based on the gained insight we design a methodology for the identification of 802.11 wireless clients and test it on two data sets of packet-level traces collected in different networks. Our results demonstrate that wireless identification is highly precise in the presence of a sufficient traffic sample.