Free Trial

Safari Books Online is a digital library providing on-demand subscription access to thousands of learning resources.

Share this Page URL

30.7 Small World VoIP-P2PSIP-Based on Trust > 30.7.3 Join and Leave - Pg. 786

786 CHAPTER 30 Security Issues in VoIP Telecommunication Networks the hub nodes are needed to transfer control messages, the hub nodes have a high likelihood of being congested, because all TCP connection requests cause the majority of the delay. Here, the proposed algorithm uses a formula to calcu- late the next best node for routing which is used in case a node is found to be affected by con- gestion. The users inside find it by checking the flag FConn which is set to one if it is found to be affected by congestion. So, the users will cal- culate the next best node for routing its service requests. the authentication is to make sure that these crawlers are not used for malicious attacks on the networks. Apart from collecting call logs from the usage, these crawlers also check the status, i.e., if the user had a new SID or a new IP, of the user using a breadth-first search algorithm. 30.7.2. Closeness-Based Routing The Figure 30-20 depicts the general routing in a cluster system. Clustering uses three algorithms. The first looks up the user in the cluster. It is done by closeness-based routing (illustrated in Figure 30-21). In case of a caller agent calling a callee, the caller needs to know the address infor- mation of the callee. The DHT uses a tuple con- sisting of key and value but this algorithm differs in a way where it uses a hash value of IP and port as the key and uses the same method to find the user object and to find the closest user for that key. It also shows how hubs act as supervisors of clusters. Hubs with higher popularity make them more trusted by users. So, each user can choose its hub based on a credibility value given to it and from the list of hubs his/her friends have. 30.7.5. Experimental Evaluation All algorithms were implemented and evalu- ated. Simulations were conducted by using real- life call logs of 97 users using mobile phones for 9 months at MIT by the Reality Mining project group. The mobile phones used were Nokia 6600s loaded with software needed for the research. The deployed test bed had around 1000+ nodes and 300+ PlanetLab machines using OpenVoIP developed by a research group at Columbia University. Setting up to six differ- ent time periods like weekday, weekend, school open winter break, daytime, and nighttime as parameters they have generated traffic on the net- work. The evaluation of the test bed showed some important things. It was found that 10% of the users were always active. The SW-VoIP shows a higher percentage of 1 or 2 hops and a lesser percentage of multi-hops than OpenVoIP. It also showed if the number of clusters was 2­4, SW-VoIP showed better results, thereby indicat- ing that the more closed the group is, the better the performance is. Further results of the evalua- tion are discussed. We observed that there are a larger number of intra-cluster calls than inter-cluster calls. We generate these two kinds of user lookup traffic on four clusters of the SW-VoIP system, respec- tively, and illustrate the system performance in the following figures. For intra-cluster calls, most of the routings are processed in 13 hops with the delay of 0­400 ms, whereas inter-cluster rout- ings have the scattered test results ranging from 1 to 6 hops with 0­1000 ms delays. 30.7.3. Join and Leave The second algorithm, described in Figure 30-22, defines the things to do by the user to join or leave a cluster. To make sure malicious elements do not join the network, it has been communicated and recommended to a certain number of members who are already on the SW-VoIP network. After this step, the user joins the network by setting up his status flag to 1 and then can actively partici- pate in the network. For a normal user to leave the process is simple. It just informs its friends and passes its stored keys to a close friend and leaves. But when it comes to a user who was a hub, the next hub needs to be calculated before the current hub user leaves. This is done by cal- culating the next popular user in the cluster and the users elect the next hub. 30.7.4. Congestion Avoidance The third algorithm is the congestion avoidance algorithm (as described in Figure 30-23). Because