Source Identification And Control For Propagation Of New-type Network Risks Based On Partial Observation Information

With the rapid development of computer technology and modern information communication technology,various online network systems and related derivative applications emerge in an endless stream.The Internet,OSNs(Online Social Networks),mobile communication networks,and wireless sensor networks are typical applications.These networks greatly change people's practical activities and living lifestyle and accelerate dissemination of multifarious online information.However,unfortunately,the development of modern network systems also has its disadvantages.The openness of network systems enables risks(e.g.,computer worm,rumors,etc.)diffusion in a very short time.In October,2016,DNS service provider Dyn suffered massive DDoS attacks,eventually resulting in nearly half of the U.S.networks failure.In addition,in recent years,the common malicious information like e-mail virus etc.,are endless and very hard to prohibit.2015 China Internet Security White Paper,points out that malicious network information will be key remediation objects of future Internet security area.In this paper,rumors of mass forwarding in online social networks,contagious worm viruses in wireless sensor networks,malicious codes on Internet and smart power station accidents on smart grid are all referred to as the new-type network risks.The new-type network risks are characterized by their ability to flood fast.Usually,they exploit mutual trust between the topologic nodes to spread to most nodes from a small number of risk sources.When large-scale network risks break out,it is of great significance to identify the sources of malicious information and evaluate network users' security states based on the limited prior knowledge,which can be used to combat the diffusers,repair network vulnerabilities,screen and protect users at high risk and block the spread of network risk in a timely manner.However,due to cost control and user privacy protection requirements,we can only detect some user nodes and can only get limited part of the observation information.Therefore,it is an important research direction to deal with the new network risk based on limited observation data.In the context thatnetwork security has become more and more severe and a growing worldwide awareness of the need for strengthened privacy protection,this paper is of great theoretical and practical value and is worthy of in-depth study.