Sybil
attacks are the fundamental problem in peer-topeer
and other distributed systems. In a Sybil
attack, a malicious attacker creates multiple fake
identities to influence the working of systems
that depend on open membership, such as recommendation
and delivery systems. Recently, a
number of social network-based schemes, such
as SybilGuard, Sybillimit, SybilInfer, and SumUp,
have been proposed to mitigate Sybil attacks.
Viswanath et al. [19] develop a deep understanding
of these approaches. It shows that existing
Sybil defense schemes, which can be viewed as
graph partitioning algorithms, work by identifying
local communities (i.e., clusters of nodes
more tightly knit than the rest of the graph)
around a trusted node. Therefore, the substantial
amount of prior research on general community
detection algorithms can be used to design
effective and novel Sybil defense schemes.
Usually, binary Sybil/non-Sybil classifiers have
high false positives; thus, manual inspection
needs to be involved in the decision process for
suspending an account. SybilRank [20] aims to
efficiently derive a Sybil-likelihood ranking; only
the most suspicious accounts need to be inspected
manually. It is based on efficiently computable
early-terminated RWs and is suitable for
parallel implementation on a framework such as
Map Reduce, uncovering Sybils in OSNs with
millions of accounts. SybilRank is deployed and
tested in the operation center of Tuenti, which is
the largest OSN in Spain with 11 million users.
Almost 100 and 90 percent of the 50K and 200K
accounts, which SybilRank regards as the most
suspicious, are indeed fake. In contrast, the hit
rate of the current user-report-based approach is
only 5 percent. Thus, SybilRank represents a significant
step toward practical Sybil defense.