| Summary: | 碩士 === 國立臺灣科技大學 === 資訊工程系 === 101 === In this study, we conducted experiments on different kinds of mobile devices, with each a set of 10 different configurations was applied and tested, to find out the minimum time to measure clock skews of enough precision so as to identify mobile devices via network connections. Clock skews of these mobile devices, or the clients, was measured by the server which has collected thousands of timestamps from the clients.
Three algorithms including linear regression method (LRM), quick piecewise minimum (QPM), and linear programming method (LPM) were compared by their performance on stability, noise effect, and vulnerability to outliers. In preliminary research part, we at first discussed hardware configuration may affects the oscillator frequency, and how these two factors would affect the measurement of clock skew. Secondly, we discussed how the rate of packets missing, standard deviation, and the ratio of packet interval in client and in server can help to judge experiment results. Thirdly, we argue that median is better to find a representative clock skew in every experiment than average and trimmed mean. And last, we summarized the difference of get time functions of different operating systems and programming languages.
According to the results of preliminary research part, we selected four parameters and designed experiments accordingly. The selected parameters include amount of packets, packet interval time, network type, and built-in operating systems of mobile devices. The experiment results suggest that larger amount of packets and longer packets interval time would derive better estimating result; Estimation would be stabler in WiFi environment than in 3G; Devices with different operating systems would perform differently in the same configuration and thus require different parameters to obtain more precise estimation. In conclusion, there is at least 98% probability that a clock skew of error from 0.7 to 1 ppm can be measured with 5,000 timestamps, 500 ms packet sending interval, from any mobile device with WiFi or 3G connection.
|
|---|
