Several Information Security Problems On Complex Networks

Many complex systems in the real world can be represented by a network, in which nodes denote individual objects of interest, and edges denote the relations between the objects. In the past, random networks have been thought as the best network model for describing real-world complex systems for many years. With the advent of information era, people can easily obtain huge amount of network data and analyze them using sophisticated computing tools, which leads to two extremely important breakthroughs in network research: small-world networks and scale-free networks. Since then, complex networks have attracted research attention of domestic and international scientists from multiple disciplines, and become hot research topics in physics, mathematics, information science, biology as well as sociology. Many systems in the field of information science and technology can also be represented by networks. For example, Internet is a network in which nodes are computers or routers and edges are communication media (e.g. optical fiber cables). World Wide Web can be represented by a network in which web pages are nodes and hyperlinks connecting web pages are edges. Wireless sensor networks are networks with sensors as nodes and message transmission relations between sensors as edges. The networks corresponding to such systems are called as information technology networks. Complex network theory and methods are providing unprecedented opportunities for studying these systems. Information security aims to protect hardware, software in information networks as well as the data in their systems from intentional attack and damage. Research topics in information security that can utilize complex network theory and methods include virus and worm propagation and immunization in computer networks, attack and defense of Internet, spam filtering based on Email networks, etc.In this thesis, after discussing complex networks and various information technology networks, we mainly conduct systematical research on attack and defense in complex information networks, modular structural analysis of Email networks, as well as virus propagation and immunization on Email networks. The research results include:(1) We propose a new set of network attack and defense strategies based on bottleneck nodes and conduct simulation experiments on general scale-free networks, modular scale-free networks and two real internet networks. The results show that the new defined attack strategies can figure out the importance of nodes and the fragility of networks better than classical hub attack. In all these kinds of networks, hub-bottleneck nodes, rather than hubs as people thought, are the most important. Conversely, attack on hub-nonbottlenecks does not affect network efficiency. Experimental results also reveal important roles of bottleneck nodes in modular scale-free networks. Theoretically, these conclusions enrich the knowledge of hubs in complex networks. Practically, they provide useful insights into how to protect various information technology systems such as computer networks and communication networks. (2) We introduce an information-theoretic clustering method for detecting modular structure of Email networks and compare the effects of different network representations on this method. Simulation experimental results on simulated networks, a real network with known modular structure and real Email networks show that this method is robust to different representations of network data, has higher accuracy for clustering Emails users than existing methods, and can better reveal the self-similarity of Email networks. (3) We develop a new model of virus propagation on Email networks. Unlike existing models, the new model is a discrete one and considers the disappearance of virus as well as dynamic immunization. Thus, the virus propagation process that the new model gives is more close to practical cases.