Congestion-Aware Path Selection for Tor
Congestion-Aware Path Selection for Tor
复制标题
DOI：
10.1007/978-3-642-32946-3_9
复制
发表时间：
2012-02
期刊：
影响因子：
--
通讯作者：
Tao Wang;Kevin S. Bauer;Clara Forero;I. Goldberg
中科院分区：
文献类型：
--
作者：
Tao Wang;Kevin S. Bauer;Clara Forero;I. Goldberg
Tor是一个由志愿者节点组成的匿名网络，它使用节点的估计带宽作为其路径选择算法的中心特征。然而，在该算法中不考虑节点上的当前负载，并且我们观察到一些节点持续处于未充分利用或拥塞状态。这可能会降低网络的性能，阻碍Tor的采用，从而减少Tor匿名集的大小。为了减少拥塞和提高负载均衡，我们提出了一个感知拥塞的路径选择算法。使用延迟作为拥塞的指标，客户端使用机会和轻量级的主动测量来评估节点的拥塞状态，并拒绝出现拥塞的节点。通过在实时Tor网络上进行的实验，我们验证了我们的假设，即客户端可以使用延迟来推断拥塞，并表明感知拥塞的路径选择可以提高性能。
Tor, an anonymity network formed by volunteer nodes, uses the estimated bandwidth of the nodes as a central feature of its path selection algorithm. The current load on nodes is not considered in this algorithm, however, and we observe that some nodes persist in being under-utilized or congested. This can degrade the network’s performance, discourage Tor adoption, and consequently reduce the size of Tor’s anonymity set. In an effort to reduce congestion and improve load balancing, we propose a congestion-aware path selection algorithm. Using latency as an indicator of congestion, clients use opportunistic and lightweight active measurements to evaluate the congestion state of nodes, and reject nodes that appear congested. Through experiments conducted on the live Tor network, we verify our hypothesis that clients can infer congestion using latency and show that congestion-aware path selection can improve performance.