Website fingerprinting and identification using ordered feature sequences

摘要: We consider website fingerprinting over encrypted and prox-ied channel. It has been shown that information on packet sizes is sufficient to achieve good identification accuracy. Recently, traffic morphing [1] was proposed thwart by changing the size distribution so as mimic some other website, while minimizing bandwidth overhead. In this paper, we point out ordering information, though noisy, can be utilized enhance fingerprinting. addition, traces of remain even under they extracted for identification. When web access performed OpenSSH 2000 profiled websites, accuracy our scheme reaches 81%, which 11% better than Liberatore Levine's presented in CCS'06 [2]. are able identify 78% morphed among websites identifies only 52%. Our analysis suggests an effective countermeasure should not hide distribution, but also aggressively remove information.