Cross-path inference attacks on multipath TCP

Abstract

Multipath TCP (MPTCP) allows the concurrent use of multiple paths between two end points, and as such holds great promise for improving application performance. However, in this paper, we report a newly discovered class of attacks on MPTCP that may jeopardize and hamper its wide-scale adoption. The attacks stem from the interdependence between the multiple subflows in an MPTCP connection. MPTCP congestion control algorithms are designed to achieve resource pooling and fairness with single-path TCP users at shared bottlenecks. Therefore, multiple MPTCP subflows are inherently coupled with each other, resulting in potential side-channels that can be exploited to infer cross-path properties. In particular, an ISP monitoring one or more paths used by an MPTCP connection can infer sensitive and proprietary information (e.g., level of network congestion, end-to-end TCP throughput, packet loss, network delay) about its competitors. Since the side-channel information enabled by the coupling among the subflows in an MPTCP connection results directly from the design goals of MPTCP congestion control algorithms, it is not obvious how to circumvent this attack easily. We believe our findings provide insights that can be used to guide future security-related research on MPTCP and other similar multipath extensions.