On the establishment of distinct identities in overlay networks

Abstract

We study ways to restrict or prevent the damage that can be caused in a peer-to-peer network by corrupt entities creating multiple pseudonyms. We show that it is possible to remotely issue certificates that can be used to test the distinctness of identities. Our certification protocols are based on geometric techniques that establish location information in a fault-tolerant and distributed fashion. They do not rely on a centralized certifying authority or infrastructure that has direct knowledge of entities in the system, and work in Euclidean or spherical geometry of arbitrary dimension. They tolerate corrupt entities, including corrupt certifiers, collusion by either certification applicants or certifiers, and either a broadcast or point-to-point message model.

Detecting Sybil Nodes in Static and Dynamic Networks

“…Sastry et al [18] presented an early version of a protocol that verifies node locations and bases node identities on their location information. Similar ideas were also proposed by Waters and Felten [20], Bazzi and Konjevod [2] andČapkun and Hubaux [4,5]. In the most general form, these works are based on measuring the delays incurred by communications between nodes and imposing a geometric structure on the "distance space" to determine node locations.…”

“…Remarkably, this intuition, while it appears key in the mobile setting, is of little use in the static case. Thus, it forms a complement to the location-and-delay based certification mechanisms of Bazzi and Konjevod [2] which were limited to the static setting. In fact, the basic observation that a pair of independent users of a mobile ad-hoc network will frequently be seen simultaneously in different locations in the network, while a Sybil node is never seen far from its "master", forms the foundation for this research as well.…”

“…Thus, a message may be signed by a node, but the identity claimed by the signing node cannot be verified directly unless we already know that node. This assumption is used instead of any specific other about the geometry (a metric, in particular Euclidean space, was the key to the design of Sybil attack detection protocols in [2]) or the topology (a fast-mixing network was crucial to the protocols in [23,22]). In other words, The physical positions of the nodes are irrelevant for our algorithm from the local network point of view.…”

“…Note that, similar to [2], an assumption is made that identifies a property of a Sybil attacker and its collection of fake identities that, to an observer capable of surveying all the events in the network, would make it obvious that is something strange going on there. In the static setting of [2], the observer would notice many identities at the exact same physical location in the network.…”

“…In the static setting of [2], the observer would notice many identities at the exact same physical location in the network. In our current dynamic setting, the image would be even more blatant: the Sybil nodes will appear interconnected in a densest subgraph or in a strongly connected component, or in both.…”