物联网恶意代码传播模型研究_英文_

物联网恶意代码传播模型研究_英文_ Abstract: Nowaday,sthe main communication ob- Key words: IoT; IPv6; worm rpopagation; worm ject of Internet is human-human, But it is foresee- model; weighted complex network able that in the near future aonbyje ct will have au - nique identification and can baedd ressed and con- I, INTRODUCTION nected, The Internet will expand to thIen ternet of Things, IPv6 is the cornerstone of Ithne ter net of The Internet of things,also known as thIen ternet of Things, In this paper,we investigate a fast aticve objects,refers to the networked inte rconnection of worm,referred to astop ological worm,which can everyday objects, It is described as a s elf- propagate twice to more than threetim es fastert han configuring wireless network ofse nsors whosepu r- a traditional scan-basedwo rm, Topological worm posew ould be toin terconnect all things, The secur- spreads over A S-level network to pology, making ity of the In ternet of t hings is still a challenge,traditional epidemic models invalid for modeling the Onceit is attacked,innumerable property admages propagation of it, For this reason,we studyto polog- will be caused, IPv6 is the cornerstone of Itnhtee r- ical worm rpopagation relying on simulations, First, net ofT hings, Hence,it is urgent and necesstaryo we propose a new ocmplex weighted network omdel, researcha ll kinds of escurity problems related towhich represents threea l IPv6 AS-level networkto - IPv6, Malicious code as used Inbtye rnet wormsi s a pology, And then,a new wormr oppagation model main way to conduacttt ack, based on thee iwghted network omdel is construc- The first known worm was tMhoe rr is worm inted,which describes the topological worm rpopaga- 1988, Snce then,new worms appearedrequ fenty,iltion overA S-level networkto pology, The simulation especay in the ast severa years, A computer illllresults verify the topological worm model and dem- worm s macous sef-propagatng code, Oncebu r- iliili onstrate the effect of parameters orno ptahgeat pon,i rowedi nto a susceptible system,it has the optential to infect manyv ulnerable hosts on thInet ernet in aspead,befoe switching to a pemutation scan rrr once thekno wn neighbors are exhausted, Yookvery shorttim e before human countermeastaukrese place, The heavysc anning traffic generatedb y the S, H, ,12, studied the emeging netwok withrr infected hostsw ill lead to network co ngestion andnon-binary connectivity, He introduced two weighted network models,WSF and WE, Alain equipment invalidation, In 2001, the Code R ed Baat ,13, poposed anothe weighted net- rrrr and N imda worms uqickly infected hundredsof wok model,called BBV, It usest he stength to rr thousands of computers,1 , ,causing millionsof describe the weighted value, dollars loss to ours ociety, The Slammer worma p- Both the Peer-to-Peer network and the AS- peared on January 2,5200th3,and uqickly spread level Internet are huge complex networks, It is throughout the Internet ,2,, Since then,the secur- quite difficult to model the worm propagation in ity threats causedby network worms haveincr eased such environment in details, Reference ,14, dramatically, Today,our computing infrastructure used them ean-field to analyze the spread of In- has much morevul nerabilities ,3,, In the epidemiology area,researchers havestu d- ternetworms, and obtained several conclu- ied the modelling and analysis of thep ropagation ofIn fact,it is not proper to model the sions, viruses for al ong time, Kephart,White and Chesspropagation through simple scale freeworm have done sa eri es of tsudies on viral infection network topology, A weighted network is a bet- ter choice, The edgeo f the graph can represent based onp iedemiology models,4-6,, Wang eta l, costs,bandwidth and so on, In this paper,we modeled virus propagation by considering virus propose a new worm model based on weighted infection delay and user vigilance, Chen et al, network topology in IPv6 network, pesented a discete-time wom model that con- rrrRecent research has shown that the noded e- siders the patching and cleaning effect during gree of AS-level Internet topology exhibits pow- wom popagation, Zou et al, pesented atwo- rrr“er law,and the topology has a great impact on factor”worm model that considers the effecto f somea pplications that run on the Internet, Ref- human countermeasures and the congestions erence ,15, showst hat the topology generated causedb y worm scan traffic,7-9,, Zou et al, by the powe-law topology geneatos esembles rrrr the AS-level Internet topology better than those explored the possibility of monitoring Internet produced by random graph generators and raffic with small-sized address spacean d pre- tstructural generators, On the other hand,so dicting the w orm propagation, Wang et al,fa studies of epidemic spead ust focuso n un- rrjpesented simulation studies of a simple vius rrweighted SF networks,and a detailed inspection popagation on clusteed and tee-like hiea- rrrrrof epidemic spreading process in weighted SF chical networks, Based o n the eigenvalues ofnetworks is still missing while real networks, netwok gaphs,Wang et al, pesented the epi-rrrsuch aspopu lation and Internet,are obviously demic threshold for virus propagation on arbi-scale-fee and with links' weights that denotef a- r trary network topologies ,10-11,,miliarity betweent wo individuals,or the time a The Internet worm spreading by network to- worm spreading from one route to another ( like pology infomation is called topological wom,t rrI people or computers) ,I n th is paper,we intend usesi nformation contained in the victim ma- to povide a detailed analysis of wom popaga- rrrchine or routers to select new targets, Email tion in weighted SF networks, worms have usedt his tactic since their incep- The remainder of this paper is organized as fol- ton,as they coect addesses fom the vctms illrririilows: we first survey differential equation models in order to find new potential targets,as did the and weighted network topology in Section II and Morris worm, Many future active worms could easy apply these technques dung the nta iliriiiil 80 2011, 1 RESEARCH PRPER 论文集锦 ds( t) Section ; Section intoduces ou poposed IIIIV rrr , = ) ks ( t) ( t) λΘ k dtweighted AS-level model of IPv6 network; Section , , di( t) V illustrates the TWM6 worm model; Section VI ( 2) = )i ( t) + ks ( t) ( t) λΘ, kk dtgives coesponding simulation esults; we con- rrr, dr( t) , cude the pape in Secon ,lrtiVII= i( t) , dt , DSCUSSO OF DFFIIIINIERENTIAL( n )1 ) P( n) i( t)? n n ( t)= ( 3) Θ QUO ODSEATINMEL E,k, any gven i The factor is theprobability that“Θ n S mode ,16-18,,each host stayins one of IIlpoints to ani nfected host”,4,, is derivedlink Θ the two states: susceptible o infectious, The r based on thceo nclusion that the rpobability a link mode assumestha t the systems homogeneouslipoints to ans -degree nodei s proportional to — each host has theeq ua pobaby to conaclrilittt , Θ Apparently,both of them odels are suitable any other hosts in the Internet, Once a host is for infected by a worm,it remains in the infectious the epidemic spread, Nevertheless,they are n otstate foeve, S mode aes the emovarrIRltkrl proper for the worm rop apgation, From variableprocess of infectious hosts into consideration assumes that thadej a-( t) ,we knowm odel ( 3) Θ ,19,, It assumestha t during the propagation of cent nodes of oinnfeec ted hosti s uniformly distrib- uted,whch s not the casine the rea AS envron- iiliInternet worms, some h osts can be e ither ment, In Ref. ,14,,Zou analyzed the deficience patched o closed,and thus they ae immune to rr of SIR model in simulating the email-virus propaga- the wom foeve, Fome wom model neglects rrrrrr tion, We use as imple two-dimensional grid net- the dynamic effect, such as huma n counter- work,shown in Figure 1,as ane xample to illus- measues on wom behavo and the changeo f rrir trate the omdeling problem, Zou pointed out that infection ate duing infection, The two-factorrr worm model considers the two factors. ( t) overestimates the effective infection links,Θ ee,we use mean-fed method to dscuss Hrili Suppose that no5 deis the initial infected node and the epidemic spread in different networks,it nfects three out of fonuer ghborng nodes at me iiii tick later ( labeled as black node) ,Th e ep idemic 1) SIR model in homogeneouns etwork We denote i( t ) as the faction of infectedr has 10 effective infection links, However,if the k hosts in the k-degree hosts et, The infection ratei sfour infected nodes are scattereind the networaks mpcty assumed iliildenoted by which is the probability that as us- ,λ ceptible nodei s infected by one neighboring infec- ted node wthn a unit tme, iii ds( t) , ) = λ , k , i( t) s( t) dt, , di( t) ( 1) ,k ,i ( t) s( t)= )i ( t) λ , dt, dr( t) , = i( t) , dt 2) SIR model in complex network Suppose,in a topological network,P( k) is the fraction of nodes that have degreek ,th e average degree of the network is E,k,= kP( k) ? k Fig, 1 Shortcomings of mean-field method 2011, 1 81 AS network by an duirnected graph by model ( 3) ,the epidemic would have1 6 effec- G= ,V ,E ,, tive infection links, Therefore Zousai d that modelv ,v denotes each ,AaSnd e (= u ,v) V ? ? ( 3) overestimates the epidemic propagation representstw oE,u,v VAS u and vw hich are? speed,On the otherh and,if we saym odel ( 3 ) is connected itwh each other, A worm can roppagateun- derstandable in modeling disease,it is really across edges, The weight value of each edge canre- not a proper model to research In ternet worm present abndwidth,time delay or something else, propaga- tion, An Internet wormi s malicious code A, Internet topology( standa- lone or file-infecting ) that propagates In Ref. ,15,,the results show that none of ttoh-e over an et- work,with or without human assistance, It can in- pology generators producestopo logies close to that fect remote hosbtsy its IP address through of theI nternet if the clustering coefficient is taken routers,The important difference between worm as the ricterion, SoT ian Bu proposed a newto polo- propagation gy generator that usesge naer alized linear prefer- and disease spreading is that a worm caninfe ct notencem odel, The new generatorallo ws us top ro- only its adjacent neighbors,but also remote hostsin ducet opologies with clustering coefficients closer todifferent sub-networks, In Figure 1,according to that of theIn ternet,mean-field theory,infected node5 can only infect The AS-level Internet topology exhibits not only node0 ,2,3 and 4, Actually,node 5 can infect power l aw, but also clustering phenomena, Re- takes farle ss edgesin tonetwor,ksuch as1 , ( t) Θany other nodites can reach through rouinters th e cently,more and more researchsestud ied the pow- consideration, er-law topology generator,sbut theyd id not take clustering phenomenian to consideration, It is use- III, DISCUSSION OF WEIGHTED NET- ful to research a newtop ology generator hwich can WORK TOPOLOGY show the powleraw and clustering phenomena tofhe The complex networki s widely applied in differentgeneratedg raph, Fortunately,Generalized Linear areas, It describes the relationship of individuals in Preference( GLP) appeared, the system and tchoell ective acts of the B, Power lawssystem,Co mplex networki s the most common Recently,the ubiquity of power-law degree idstri- abstraction form,and mancyo mplex systems can be bution in real-life networks has attractedlot ao f at- abstracted as complex network, Many social, tention, Examples of such networ(k ss cale-free net- biological,and communication systems can be works or SF networks s hfor ort ) are n umerous: properly described as complex networks iwth vertices representing indi- viduals or organizations and links thesei nclude the Internet,the World Wide Web, mimicking the in- teractions among social networks ofa cquaintance or other re lations them, Howeve,rso fars tudies on epidemic spread between in dividuals, metabolic networks, integer just focus onu nweighted SF net- works,and a network,sfood webs,etc, The ultimate goal of the detailed inspection of epidemic sprea- ding process in study onto pological structure of networisk s to un- weighted SF networkiss still missing while real derstand anedx plain the working of systembus ilt network,ssuch as oppulation and Inter- net, are upon those networ,kfor si nstance,to understand obviousy scae-free and with inks weights lll’ how the topology of the W orld Wide Web affects that denotea mfiliarity between twiond ivid- uals ( like people or computers ) , One can ea sily take Web surfing and searchen gines,how the structure cognzance of how then ks' weghts affectt he iliiof scoial networks affects the spreadis eaofs eds,in- epidemic spreading process, formation,rumors or othert hings,how the structure We represent theop tology of thel ogical Internetof a food web affects population dynamics,and 82 2011, 1 RESEARCH PRPER 论文集锦 'ωji ' so no,where represents a sum over the = 1,{ i} m ? ' , , i The algorithm of GLP is: our model captures existing nodes towh ich the new nodj eis connected,two events co rresponding to the a ddition of a 2 ) BBV model improves the WSF in three as- new node and thadde ition of al ink, We start pects, Frst,the weghts of the network are ob- iiwith m nodes connected throumg h) 1 edges, 0 0 and at each tm estep we peom one of the i-rfrtained by vertex strengths, s= ,Sec-ω ?i i ij jV( i) ?following two operations: ond,the weghts of edges chandgeyn amcay in the iill1 ) W ith probability p ,we addm mnew ? 0 process of network ev outon, The agorthm s as liliilinks, For each end of eaclinhk ,node i is cho- follows: sen w ith probabilityas defined in Eq.d ?Startng from as manumber of vertces,iilli ? m 0 ( 4 ) ,T his in corporates the fact thante w linksand , which is assigned with initialm)1 edges 0 preferentially connect oppular nodes,weight ,ω 0 4 )d =/ ( d ( d ) )) At each tme step we add a new wnhodch e )( ii? β β? ? i i j jlinks to me xisting nodesin the system, The proba- 2 ) With probability 1 node,we add a new )p bility that a new nodj we ill connect to anex isting The new node hmas new links, Each link is nodei is connectedt o node i aeady pesen in he lrrtt s isystem w ith pobaby d as defned inriliti i ?( 7) ? s j iEq. ( 4 ) , ? j C, Existing weighted complex network weight of each new edisge fix ed to av alueThe ? Typical weighted networkt opology models are WSFω ,Moreove ,rthe presence of a new edillge in w- 0 ( weighted scale free) and BBV( Barrat-Barthelemy- troducev ariations of the xeisting weights across theVespignani) ,I n the nexts ection,we will illustrate network, In particular,the local rearrangemenotsf the two omdels, weights between and its neighbor i j V( i)accord-? 1) WSF mode derved from scae-free network,lil ing to thes imple rule + ,where ωωΔωΔω? ij j ij j iiforms network top oogy structure a ccordng toIt li = ω ijpreference emchanism, In the process of evolution,the new nojd es connected,iδ s a weghted vaue s set to eachd gee, The weghtediliii value of edgesis fixed,and it does not chanwgeith the evouton of networks tructure, The agorthm s, WORM PROPAGATION MODEL BSEDliliiIVA as follows: ON NETWORK TOPOLOGYStarting from as mall numberm of vertices.? 0 At each time step we add a new wnhoidch e ?For many observescda e-free networks wth ,, 2,li γ links to me xisting nodesin the system, The proba- thesem odels fit their degree exponent featurveer y bility that a new nodj we ill connect to anex isting well, Howeve,ra major challenge arises whenu sing nodei is thesem odels to reproduce thIPev 6 Internet topology k iwith quite a small degree exponent because mof ost ( 5) =? ik them have them taton of , 2 ,7,, liiiγ j? jReference, 21,proposed a newAS -eve topoo- lllAssign a weight to then ewy estabished ink lll? gy model of IPv6 network, Here,we propose a new asji weighted AS network Iof Pv6 Network( WASTv6) , k i =( 6) ω ji k 'i1 ) We startw ith fully connected graph of m 0? ' , , nodes and at eachtime step we perform oneth eof i 2011, 183 model to study the spread of susceptible-in-BBV following threeo perations: fectious ( SI) worm model, The individuals can 2 ) With probability p,we add (m m ,m )new 0 be divided into two states,either susceptible or links: we r andomly select a node as thste art ing infected, The host is represented by a vertex of point of the nelwin k,while the other end of tlhinek the network and the links are the connections be- is selected with probability given in Eq. ( 8) , tween hosts, They define the infection transmis- 3 ) With probability q ,we rewire m links: we sion by the spreading rate randomly select a node and linak connected to α ω ijit,Next we removethi s link and replace it with a = , λ, 0( 9) α ij ， ， ω Mnew link that connects ithw a new node chowsienth In this paper, we propose a ball-like model probability given in Eq. ( 8) , named t opological worm model in IPv6 network4 ) With probability 1 ) p q) , w e add an ew ( TWM6) ,Th e ustraton of theTW M6 s as fo- illiilnode,which has m newli nks with probability given lows: in Eq, ( 8 ) connected to nodlresea dya presentin The ball-like model is derived from theW ASTv6 the system,model, Any two vertices select a patha ccording to 5) The weight of each new edisge fix ed to av aluecertain strategy ( depth irfst search) from the ω 0WASTv6 model to compose a new ined gethe ball- )like model, The weight of the new edisge e qual to + kkε i=( 8) ? ) the sum of thee igwhts of thes elected path, The en- ij + kε? tire operation is not ifnished until any twov ertices j have edges ancdorr esponding weights,except outli- This model not only matches theo wper-law expo- er nodes, We start byse lecting one vertexra ndomly nent of thAe S-level of IPv6 network,but also takes and assumite is infected, Routing worm will spread the clustering behavior of the Internet into in the networkins accordance with the rule of account,Moreove ,rthe way ofass igning weight to Eq, ( 9) , edges is in accordance with the real Internet, That is at the V, SMULATON AND ANALYSSIIIIAS-level of IPv6 network,s the new es tablished edge between two ASes usually affects the tra ffic We simulate the propagation of routing wormin dis-between the two ASes and aAssSesocia ted with the cretet ime based on thea lbl-like model, The system two, in our simulation consists of N hosts that canrea ch V, BALL-LIKE WORM MODEL other hostascc ording to the abll-like graph, A host staysin one of the three states att imaney: suscep- General epidemic dynamics researchis mainly con-tible,infectious,or immune, A host is in im- “cerned iwth disease and social systems,which are mune”state when it is immunized, Each copoyf not the same as worm propagation in the the worm on aninf ectious host canin fect otherh osts Internet,Th e individuals in the disease and social according to the networktop ology, systems n caonly interact with their adjacent Figure 2 shows the roppagation of TWM6 worm,W e can see that the wsormprea ds extremely individuals, On the fast, contrar,ythe worm cainn fect not only its neighbors, When t = 25, TWM6 finishes its first spreading but also compromise remote hosts througroh ute rs, stage,and entersn to the seconsdp readng stage, iiwhich is not the cainse disease spread, So it is not It infects 80% of vulnerable hosts whent i s aboutaccurate to omdel the wormr oppagation,which only 62, We know that despite the hugeadd ress spaceo ftakesa djacent interaction into consideration, IPv6,TWM6 can still forml arge scale propagation, R, Pastor-Satorras eat l. took advantage o f 84 2011, 1 RESEARCH PRPER 论文集锦 weghted power-aw networks under the sacomnde - ili tion, Fig, 2 Topological Worm Model in IPv6 network Fgure 3 shows the effects of parammeteron thei 0 TWM6 worm rpopagation, From thei gfure,we can Fig, 4 Effect of parameter mon the propagation of TWM6 0 see that thien itial number ofve rtices have impact Figure 5shows the effect of parameteron theε on theTW M6 worm rpopagation, Whenm = 5,the 0 TWM6 worm rpopagation, The parametercan ad-ε TWM6 spreads faster than the case wmhenis 10 or 0 just the node icpk-up probability higher or lower 15,It is better to choose a promperto satisfy cer- 0 than the purelin ear preferential attachmenint the tain propagation scene,BA model, From thei gfure,we can see that the higher is,thefaster theTW M6 worm spreads, ε Hence,the parametercan be a key ctoon trolling ε the propagation speed ofTW M6 worm, Fig, 3 Effect of paamete mon the popagation of TWM6rrr 0 Figure 4 shows the eff ectof onouting wormrδ propagation, From thei gfure,we can see thatro u- ting worms preads more uqickly while is smaller, δ Fig, 5 Effect of paamete mon the popagation of TWM6rrr 0 It is consistent with the conclusion,and it demmon- strates thatla rger value of induces larger disper- δ V II, CONCLUSIONSsion of weight of networks, So the fi gure tes us ll that larger dispersion of weight of networks re sults This paper idscusses aboutI nternet worms preadingin sower spreadng, That means that wormsprea ds li in weighted complex network,ssuch as opwer-lawfaster on un weighted power-law networks th an 2011, 185 ,8, DALEY D J,GANI J, Epidemic Modelling: an Introduc- network, We propose a new weighted network tion,M,, Cambridge: Cambridge University Press,1999, named WASTv6 to represenrtea tlh eA S-level In- ,9, ZOU C C,TOWSLEY D,GONG W, On theP erformance ternetn etwork, On the basis of the newwe ighted of Internet Worm Scanning Strategies,J,, Journal of Per- network,we give a novel topological worm rpopaga- formanceE valuation,2006,63( 7) : 700-723, tion model called TWM6, This model well illus- ,10, PASTOR-SATORRAS R , VESPIGNANI A, Epidemic trates the roppagation characteristics of topological Dynamics and Endemic States in Complex Networkswormin IPv6 environment, Somei mportant conclu- ,J,, Physical Review E,2001,63( 066117) : 1-8,sions are given based on theo dmel, In the ufture, ,11,PASTOR-SATORRAS R , VESPGNAN A, Epidemic IIwe will mainly focus on thdeef ending strategies a- Spreading in Scale-free Networks,J,, Physical Review gainst the TWM6 worm, Letters,2001,86( 14) : 3200-3203, ,12, YOOK S H,JEONG ,BBS A , eghted E- HARAIALWi volving Networks,J,, Physical Review Letters,2001, 86( 25) : 5835-5838, Acknowledgements ,13, BARRAT A, BARTHELEMY M, VESPIGNANI A,W eighted Evolving Networks: Coupling Topology and This work was supported by the Ministry of Education Re-Weights Dynamics,J,, Physical Review Letters,2004, search Project for Returned Talents after Studying A- 92( 22) : 228701, broad,the Ministry of Education Project of Science and ,14, ZOU C C,TOWSLEY D,GONG W, Modeling and Sim- Technology Basic Resource Data Platform ( No, 507001 ) , ulation Study of thePr opagation and Defense ofInt ernet International Scientific and Technological Cooperation Pro- Emai Worm,J,, IEEE Transactions on Dependabe and llgam ( S2010GR0902) ,and Chinese Univesities Scientific rrSecure Computing,2007,4( 2) : 105-118, Fund ( 2009RC0502) , ,15, BU T,TOWSLEY D, On Distinguishing betweenIn ternet Power LawT opology Generators,C,/ / Proceedings of the Refeencesr21st Annual Joint Conference of theIEEE Computera nd Communications Societies: Jun 23-27,2002,New York, USA, Piscataway,NJ,USA: IEEE,2002,2: 638-647, ,1, ZOU C C,TOWSLEY D,GONG W,et al, Advanced ,16, SU F,LIN Z,MA Y, Effects ofF irewa on WormP rop- llRouting Wormand ts Security Chaenges,J,, Simua- Illl agation,C,/ / Proceedings of International Conferenceon tion,2006,82( 1) : 75-85, Communications Technology and Applications: Oct 16- ,2, PORRAS P A,SAD H,YEGNESWARAN V, A Multi- II 18, 2009, Beijing, China, Piscataway, NJ, USA: perspective Analysis of the StormW orm,R,, SRI Techni- IEEE,2009: 880-884, cal Report,2007, ,17, SU F,LIN Z,MA Y, A Survey ofIn ternet Worm Propa- ,3, NAZARIO J, The Conficker Cabal Announced,EB / gation Models,C,/ / Proceedings of International Confer- OL,,Arbor Networks,2009, http: / / ence on Broadband Network Mu ,tim edia Technoogy: llasert, arbornetworks, com / Oct 18-20, 2009, Beijing, China, Piscataway, NJ, 2009 /02 / the-conficker-cabal-announced / , USA: IEEE,2009: 453-457, ,4,NAZARIO J,ANDERSON J,WASH R,et al, The Future ,18, SU F,LIN Z,MA Y, Worm Propagation Modeling Based of I nternet Worms, Blackhat Briefings,2001, http: / / on Two-factorM odel,C,/ / Proceedings of the5 th Inter- www, crimelabs, net / docs / worm, html, national Conference onW ireless Communications,Net- ,5, LI P,SALOUR M,SU X, A Survey ofn ternet Worm De- I working and Mobie Computing: Sep 24-26,2009,Bei- ltection and Containment,J,, IEEE Communications Sur- jing,China, Piscataway,NJ,USA: IEEE,2009: 1-4, veys ,T utorials,2008,10( 1) : 20-35, ,19,SU F,LIN Zn,MA Y, Modeing and Anaysis of nternet llI,6, WEAVER V P N,STANFORD S,CUNNNGHAM R, A II Worm Propagation,J,, The Journal of China Universities Taxonomy of CompuWteor rm s,C,/ / P roceedings of the of Posts andTe lecommunications,2010,17( 4) : 63-68, 2003 ACM Workshop oRna pid Malcode: Oct 27,2003, ,20, HUFFAKER B,CLAFFY K, Visualizing IPv6 AS-level Washngton,DC,USA, New York,NY,USA: ACM,2003: i nternet Topoogy 2008,EB / OL,, http: / /l 11-18,Iwww, caida,or g / research / topology / as core network / ,7, BAILEy N T, The Mathematical Theory ofIn fectious Dis- ipv6, xml, eases andIt s Applications,M,, New York: Hafner Press, 1975, 86 2011, 1