Comparison between normal and TOR-anonymized web client traffic

Abstract

TOR is known free tool for traffic anonymization on the Internet. The adverse aspect of using TOR is significant increase of overhead and decrease of the traffic speed. So far no study focusing the quantification of such decrease was published. TOR operation is based on “routing” the traffic through several nodes resulting in difficult and practically impossible direct calculation of the delay caused by TOR application. Therefore the study was performed to quantify the delay associated with using TOR comparing ordinary traffic. Www was used as a model of Internet traffic. The sample of 14 stable www pages and 8-10 files available via http throughout the world was chosen and the round-trip time was measured approx. 10 times both with TOR (at least in two configurations, first automatic and the second manually modified) and without TOR. For manual TOR measurement two onion routers were configured in two locations. Measurements were made in various days of week and various times to eliminate possible fluctuations due to varying traffic target www servers. The main result of the study is that the factor of roundtrip time increase varies between 2 and more than 100. In case of configured TOR networks the ratio is lower but still high,1.7 and 34. This is quite rough result that is shown in details in the article. The results show that TOR is useful anonymization tool but the delay caused by its use is enormous. Providing of more exact estimations would require more measurements however.