Digging into IP Flow Records with a Visual Kernel Method - Computational Intelligence in Security for Information Systems - page 101

fake content distribution, free riding (peers who do not share, but consume resources),
whitewashing (peers who leave and rejoin the system in order to avoid penalties) and
search scalability problems. Open and anonymous nature of P2P applications lead to
complete lack of accountability of the content that a peer puts in the network. The
malicious peers often use these networks to do content poisoning and to distribute
harmful programs such as Trojan Horses and viruses [2].
To combat inauthentic downloads as well as to improve search scalability, this pa-
per proposes an adaptive trust-aware scalable algorithm. The proposed scheme con-
structs an overlay of trusted peers where neighbors are selected based on their trust
ratings and content similarities. It increases search efficiency by taking advantage of
implicit semantic community structures formed as a result of topology adaptation.
The novel contribution of the work is that it combines the functionalities of trust man-
agement and semantic community formation. While the trust management scheme
segregates honest peers from malicious peers, the semantic communities adapt topol-
ogy to form cluster of peers sharing similar contents.
The rest of the paper is organized as follows. Section 2 discusses some related
work. Section 3 presents the proposed algorithm for trust management. Section 4
introduces various metrics to measure performance of the proposed algorithm, and
presents the simulation results. Finally, Section 5 concludes the paper while highlight-
ing some future scope of work.
2 Related Work
In [3], a searching mechanism is proposed that is based on discovery of trust paths
among the peers in a peer-to-peer network. A global trust model based on distance-
weighted recommendations has been proposed in [4] to quantify and evaluate the
peers in a P2P network. In [5], a protocol named adaptive peer-to-peer technologies
(APT) for the formation of adaptive topologies has been proposed to reduce spurious
file download and free riding in a peer-to-peer network. The scheme follows a defen-
sive strategy for punishment where a peer equally punishes both malicious peers as
well as neighbors through which it receives response from malicious peers. This strat-
egy is relaxed in the reciprocal capacity-based adaptive topology protocol (RC-ATP),
where a peer connects to others which have higher reciprocal capacity [6]. Reciprocal
capacity is defined based on the capacity of providing good files and of recommend-
ing source of download of the peers. While RC-ATP provides better network connec-
tivity than APT, it has a large overhead of topology adaptation.
There are some significant differences between the proposed algorithm in this
paper and APT and RC-ATP. First, in the proposed scheme, to avoid network parti-
tioning, the links in the original overlays are never deleted. Second, the robustness of
the proposed protocol in presence of malicious peers is higher than that of APT and
RC-ATP protocols. Third, as APT and RC-ATP both use flooding to locate resources,
they have poor search scalability. The proposed scheme, on the other hand, exploits
the advantages of the formation of semantic communities to improve the quality of
service (QoS) of search. Finally, unlike APT and RC-ATP, the proposed scheme
punishes malicious peers by blocking queries initiated from them.