Conclusion

In this paper, we demonstrated that a WM technique can perform reliable and accurate measurement on the 802.11 traffic, in non-ideal channel condition. We discussed how to merge traces from multiple sniffers to increase the wireless monitoring capture performance. Results from a controlled experiment, with clients having different signal conditions, show that the WM technique captures wireless side statistics with 1% and 3% error bounds at IP and MAC layers, allowing a reliable analysis of the collected traces.

We believe that the described controlled experiment presents a worst case scenario for the WM technique. This is because our Bad client case, where the client is located on the marginal line of AP coverage, is severe and unlikely to occur in a real environment. The MAC layer capture performance of the WM technique would be much better in a WLAN environment where APs are positioned so that clients in most locations of interest can find at least one AP with good signal conditions.

Based on the results in this paper, we are currently using the WM technique to analyze the WLAN traffic in a Computer Science department environment. Some initial results can be found in 12 (12) on traffic characterization and in 11 (11) on anomaly detection. Besides characterizing WLAN usage patterns, we are using the traces for multiple APs to analyze user roaming patterns, co-channel interference and interactions between different APs. In such experiments, we expect that combining wired monitoring data, such as Inter Access Point Protocol information (IEEE Std 802.11f), with WM analysis, would give better measurement capabilities, e.g., on the roaming behavior of the mobile users and the handoff process.

We believe that our results should encourage the wireless research community to use WM techniques in many research areas, including traffic analysis, user mobility and handoff analysis, and MAC/PHY anomaly detection.

0.9