Search

The paper divided the decentralized lookup algorithms in peer-to-peer networks into two classes: structured and unstructured. The lookup mechanisms, network topology and lookup efficiency of these two types of algorithms were compared. From the comparison, it is concluded that there exists tradeoff in lookup algorithms. For example, to improve the time efficiency, the algorithm has to increase the routing table size or network overhead. The paper can be a reference for the design of new lookup algorithms or application of peer-to-peer lookup algorithms in related areas such as Grid computing. [ABSTRACT FROM AUTHOR]

Based on the artificial immune system's clonal selection theory and shape space theory, this paper brought forward a novel shape-space based clonal selection algorithm. This algorithm is used to the application of multi-modal function optimization. The results indicate the proposed method is much better than other methods. [ABSTRACT FROM AUTHOR]

Anomaly detection techniques can detect novel attacks. This paper proposed two approaches based on statistical techniques. One of them is based on a maximum entropy method, which first educes the distribution of user's normal behavior, then decides the discrimination limit. The other is based on Knearest neighbor (KNN) classifier. It is probabilistic but distribution-free. The preliminary experiments with 1999 DARPA tcpdump data show that both method can effectively detect intrusive attacks. In the end, a comprehensive comparison of these two methods was given. [ABSTRACT FROM AUTHOR]

This paper sketched out the design of a decentralized, scalable, self-organizing and fault-tolerant location algorithm for peer-to-peer networks. The principle of the algorithm is as follows. A finite linear space, whose size is Mκ, is evenly divided into M parts. Again, the sub-part is recursively divided into M equal parts until all sub-parts can only contain one unit (point). Data and node's identifiers can be mapped onto the points of the linear space by using a hash function. Each node's routing table contains the location information of a small subset of other nodes, whose identifiers have some special relations to the present node's identifier in the linear space. With the routing table, one node can locate other node that is responsible for the given data items within O (log N) hops. The results from simulation experiments show that increasing the value of M can enhance the routing performance of the algorithm. In a simulation network with 104 nodes, if the value of M is 16, the average locating path length of the algorithm is only 70%/0 of that of Pastry and Tapestry. [ABSTRACT FROM AUTHOR]

This paper advanced a new algorithm to detect network intrusions using sequences of system calls. This algorithm uses a data structure called weight tree, first it uses normal system call trace to build weight tree forest, which have to be pruned periodically in order to learn new pattern and eliminate impurity. Then it scans abnormal trace using those trees and gets the corresponding weight sequences. Those weight sequences can tell us if something abnormal has happened or not. It acquired good results in experiment. [ABSTRACT FROM AUTHOR]

This paper brought forward a new immune algorithm, multi-level negative selection algorithm. Compared with Forrest's negative selection algorithm, it enhances detector generation efficiency in essence. This algorithm integrates clonal selection process into negative selection process. Based on the analysis, this algorithm was applied to network intrusion detection and has a very good result. [ABSTRACT FROM AUTHOR]