The main contribution of this paper is a content-based full- text fastsimilaritysearch algorithm, called P2PFastSS, which uses the edit distance metric and is applicable to the service discovery domain. The edit distance  is the minimum number of operations required to transform one string into another, using the operations: deletion, insertion, and replace- ment. The proposed algorithm can be used on top of any DHT, since it uses only the basic operations get and put. Thus, P2PFastSS enables a peer to search for a similar keyword in any text-based content, such as texts, service descriptions, or abstracts. In the following, the term document refers to a ser- vice or data description. It is assumed that these documents are stored randomly on peers and are indexed either by the peer storing the document or by the peer sending the document for storage. This holds especially true for a distributed service dis- covery environment in which multiple decentralized services are offered independently to various service users.
Considering different architectures of peer-to-peer sys- tems, system designers should evaluate the requirements for their particular applications and choose a topology for the platform that matches their needs. To compare them briefly (see table I), we can say that in pure peer-to-peernetworks every peer is given equal responsibility irrespec- tive of its computing/network capabilities, this can lead to reduction of performance as less capable nodes are added. Pure p2p systems lack manageability since every peer is its own controller. Unstructured pure p2p systems in which blind flooding search is used are not salable since in large scale systems the large number of exchange messages lim- its the scalability. Using structured systems or intelligent search approaches can solve scalability limitation. The disadvantage of standard structured systems is that it is hard to maintain the structure required for routing in a very transient node population, in which nodes join and leave at a high rate. It should be taken into account that some structured systems like Chord have overcome to this prob- lem and they can adapt efficiently as nodes join and leave the system.
In unstructured peer to peernetworks it is difficult to acquire the global knowledge. Which makes searching more difficult. The previous unstructured peer to peer applications used Flooding and RW are two typical examples of blind search algorithms by which query messages are sent to neighbors without any knowledge about the possible locations of the queried resources or any preference for the directions to send which causes high message cost and response delay. The proposed dynamic search in unstructured peer to peer network searches with the previous knowledge and uses the super peers in a way to reduce message traffic. A super-peer is a node in a peer-to-peer network that operates both as a server to a set of clients, and as an equal in a network of super-peers and reduces massage traffic. It combines the features of client server model and pure peer to peer model which makes search more efficient. It distributes the load among the high capacity peers or super peers.
Abstract: The huge popularity of recent Peer-to-Peer (P2P) file sharing systems has been mainly driven by the scalability of their architectures and the flexibility of their search facilities. Such systems are usually designed as Unstructured P2P networks. So, designing an efficient search algorithm is a key challenge in unstructured peer-to-peernetworks due to the unstructured paradigm. In this study, we proposed a Search Result Path Caching algorithm (SRPC). The proposed scheme combines the Dynamic Search (DS) algorithm and result path caching for effective search. This proposed algorithm takes the advantages from dynamic search and path caching technique works with the queried results. After the successful discovery of results the queries are returned to originator, according to our proposed algorithm the originator will stores the results for future references. In future stored references are used to search the information without querying the overall network. We analyze the performance of our algorithm based on some performance metrics including the success rate, search time and search efficiency. The numerical results shows that the proposed SRPC algorithm performs about 5 times better than DS, 125 times better than flooding and 275 times better than Random Walk (RW) in power-law graphs.
This paper presents a routing approach for unstructured peer-to-peernetworks that is based on query content and also connection reliability. We introduced an approach in determining peer for routing purposes at a minimal cost and efficient network load. The basic idea of this approach is to use minimal information and without the use of global knowledge in determining relevant peers to be routed. The simulation results showed that our Relevance based routing approach outperforms Most-Query Hits approach in terms of message used per query hit and also query times.
An important issue in peer-to-peernetworks is discussed in this paper. peer-to- peernetworks appeared as common method of sharing huge volume of data. These networks allow users to share their resources as completely distributed ones. Here by changing the “Chord” protocol`s structure among increasing the package` transmit speed we assure high reliability in package transmit. Simulation results showed function development of new protocol comparing to the base protocol.
ABSTRACT: Security is one of the most critical constraints for the expansion of peer-to-peernetworks. In a peer-to- peer (P2P) network, every machine plays the role of client and server at the same time. In peer-to-peernetworks, one of the most important issues is trust management. Peer-to-peernetworks rely on other peers to accomplish the tasks. Peers need to trust each other for successful operation of the system. While communicating in between peers trust formation is very important to take service from the unknown resource. In this paper, we study trust models based on various approaches like reputation, service and recommendation.
Web content is under attack by state and corporate efforts to censor it, for political and commercial reasons [9, 17, 40]. Peer-to-peernetworks are considered more robust against censorship than standard web servers . However, while it is true that many suggested peer-to-peer architectures are fairly robust against random faults, the censors can attack carefully chosen weak points in the system. For example, the Napster  file sharing system has been effectively dismembered by legal attacks on the central server. Additionally, the Gnutella  file sharing system, while specifically designed to avoid the vulnerability of a central server, is highly vulnerable to attack by removing a very small number of carefully chosen nodes [34, 35].
Many early examples of peer-to-peernetworks and protocols were never thought of in quite the same way that modern peer-to-peer systems are thought of. Even though connecting machines together not as clients and servers but as equal peers dates back to the original ARPANET, it wasn't until 1998 that an application (Napster) was written that is P2P in the modern sense (Magoules et al 2008). Between these two times, the mid 1960s and 1998, many applications were written that are quite peer-to- peer in their nature. It may seem, with hindsight, that a P2P solution was a natural fit for some problems and solutions were written without much thought given about this being a new technique in itself. This is interesting when one considers that some of the most decentralised peer-to-peer examples are actually within these early systems. Such examples are covered further in the next chapter.
3.4 Expiration and Consolidation of Reputa- tion Scores
In describing how credits for query-response processing, file transfer, and staying online are processed for both DCRC and CORC schemes, we assumed that the time-stamped rep- utation scores are sent to the respective peers immediately (except for the debits in DCRC, that are retained at the RCA). In the case of CORC, the expiration duration for the reputation scores determines when the reputation scores ex- pire. Too long a duration could mean that a peer has many encrypted reputation scores from the RCA. This could po- tentially be a problem when the peers want to present their reputation to other peers in the network. However, a long expiration duration is beneficial for peers who go off-line for a long time. This is because a short expiration may not allow them to use their earned reputation when they come back online. In the case of DCRC, the time stamp is not im- portant because of the allowance of debits to the reputation scores.
Among resource discovery techniques, BFS is the most used resource discovery technique for unstructured P2P networks being used by Napster, Gnutella, Int-BFS  and Learning Automata-based Resource Discovery (LARD). Napster and Gnutella use BFS that floods the network with queries. Int-BFS  floods the system during its initial stage but will send fewer queries once the network has learned from its previous queries. LARD uses BFS, but focus on the routing of the unstructured P2P network in order to minimise the effect of the query flooding .
In addition, as described in , Freenet uses its datastore to increase system performance. When an object is returned (forwarded) after a successful retrieval (insertion), the peer caches the object in its datastore, and passes the object to the upstream (downstream) requester which then creates a new entry in its routing table associating the object source with the requested key. So, when a new object arrives from either a new insert or a successful request, this would cause the datastore to exceed the designated size and Least Recently Used (LRU) objects are ejected in order until there is space. LRU policy is also applied to the routing table entries when the table is full. Figure 7 depicts a typical sequence of request messages. The user initiates a data request at peer A, which forwards the request to peer B, and then forwards it to peer C. Peer C is unable to contact any other peer and returns a backtracking failed request message to peer B. Peer B tries its second choice, peer E, which forwards the request to peer F, which then delivers it to peer B. Peer B detects the loop and returns a backtracking failure message. Peer F is unable to contact any other peer and backtracks one step further back to peer E. Peer E forwards the request to its second choice, peer D, which has the data. The data is returned from peer D, via peers E, B and A. The data is cached in peers E, B and A, therefore, it creates a routing short-cut for the next similar queries. This example shows the overlay suffers from security problems such as man-in-middle and Trojan attacks, and the
Once the network reaches a certain size, the servers become a bottleneck to the performance, and users can no longer search the entire network for things they are interested in (a sort of unwanted clustering).
The most recent overlay network based on MFTP, Overnet , uses the DRM algorithm to decentralize searching and publishing. Each node in the
Non-music exchanges over Napster were never more than marginal, at least compared to alternative, content-agnostic systems such as Gnutella. Some alternative Napster servers such as OpenNap which started as "safe-havens" for Napster users when Napster began filtering content, did for a while begin to fill the gap, tying together former Napster clients, clones and variations with a new kind of server that extended the original Napster protocol to all file types. No matter how much the Napster model was reengineered, however, the fundamental requirement of a "Napster-compatible" central server remained a serious constraint for a network based on this technology or any of its clones. To transcend this limitation, other protocols and architecture models are needed-for example, serverless networks in the style of Gnutella.
New questions, related to the privacy of P2P network users, were opened during our research. We have pointed out that this issue has not been sufficiently discussed, for example previously proposed easily realizable DDoS attacks, or listening queries attacks, which leads to lowering the anonymity mainly there, where users from social networks, share data, but try to stay anonymous (hiding behind nicknames). We showed that it is quite easy to collect data about communicating entities and get enough data to be able to make judgments about user’s identity and behavior.
There are increasing concerns on Peer-to-Peer (P2P) applications by Internet Service Providers (ISPs). P2P traﬃc consumes an increasingly signiﬁcant frac- tion of network bandwidth of ISPs. A recent study  estimates that the aggre- gated traﬃc of all P2P applications contributes to about 60-70% of the traﬃc in Internet and about 80% of the traﬃc in the last-mile providers’ networks. Thus P2P traﬃc increases network utilization, and can result in performance degrada- tion to other applications. More importantly, P2P traﬃc is becoming a ﬁnancial burden to the ISPs. Most ISPs connect to the Internet through their providers, and pay for the transit services. However, it may well happen that P2P clients in an ISP’s network upload a substantial amount of data to their peers in the ISP’s provider networks; hence, the ISP may become a content provider and be charged by its network provider unnecessarily. In a recent study  on Skype, the authors ﬁnd that many universities are hosting a large number of Skype su- per nodes, and becoming potential providers to the rest of the Internet. In such networks, the outbound P2P traﬃc can be 5 times more than normal inbound traﬃc . This can incur substantially higher operational cost to the university network. The world-wide extra costs due to P2P traﬃc are estimated to be in excess of e 500M per annum .
A hybrid unstructured P2P network allows the existence of infrastructure nodes, often referred to as super-peers (or super-nodes or overlay nodes). This creates a hierarchical overlay network that addresses the scaling problems on pure unstructured P2P networks such as Gnutella. A peer in such network can typically change roles over time. For example, a regular peer can become a super-peer that takes part in coordinating the P2P network structure. KaZaA ( http://www.kazaa.com/ ), which is based on the FastTrack (http://www.fasttrack.nu/) protocol, is an example of a hybrid unstructured P2P network. This network uses specially designated super- peers with high bandwidth, disk space and processing power. When a peer joins the network, it is assigned to a super-peer. The peer then informs its super-peer about the content that it will share. The super-peer facilitates the search by maintaining a database that maps content to peers, and tracks only the content of its assigned peers. Similar to the centralized design, the super-peer plays the role of a directory server, although only to its assigned peers. The super-peers together create a structured overlay of super-peers, which makes search for content more efficient. A query in this network is routed to a super-peer. Then, as in the decentralized design, the query is flooded in the overlay super-peer network. The super-peer then responds to the peer that originated the query with a list of peers having the content. The hybrid networks are no longer dedicated to a single server, since the database is distributed among the super-peers. Moreover, the size of the database is relatively small, since each super-peer tracks only the contents of its assigned peers. However, the drawback of this approach is that it is considerably complicated, and requires non-trivial maintenance of the overlay network. Moreover, the fact that super-peers may have more responsibilities than ordinary peers can result in a bottleneck.
In order to mitigate asymmetry of interest and scale to higher turnover rate, there is a need in shared history. Shared history means that every peer keeps records about all of the interactions that oc- cur in the system, regardless of whether he was directly involved in them or not. It allows players to leverage off of the experiences of others in cases of few repeat transactions. It only requires that someone has interacted with a particular player for the entire popu- lation to observe it, thus scales better to large populations and high turnovers, and also tolerates asymmetry of interest. Some examples of shared history schemes are   .
Several researchers have proposed solutions to the problem of content and index poisoning, especially in DHT-based overlay networks. For example, a scheme of maintaining a centralized database of blacklisted peers was proposed in . In addition, various reputation systems, such as peer reputation , object reputation , and hybrid reputation  were also suggested to fight against content poisoning by introducing the concept of distributed voting, which adds an extra overhead on peers as they need to collect votes and process them. Such reputation systems are also vulnerable to malicious votes and other similar attacks.