Dynamics of Malicious Software in the Internet

Slides:



Advertisements
Similar presentations
Peer-to-Peer and Social Networks Power law graphs Small world graphs.
Advertisements

Scale Free Networks.
A Distributed Algorithm for the Dead End Problem of Location Based Routing in Sensor Networks Le Zou, Mi Lu, Zixiang Xiong, Department of Electrical Engineering,
Modeling Malware Spreading Dynamics Michele Garetto (Politecnico di Torino – Italy) Weibo Gong (University of Massachusetts – Amherst – MA) Don Towsley.
RESILIENCE NOTIONS FOR SCALE-FREE NETWORKS GUNES ERCAL JOHN MATTA 1.
IEEE ICDCS, Toronto, Canada, June 2007 (LA-UR ) 1 Scale-Free Overlay Topologies with Hard Cutoffs for Unstructured Peer-to-Peer Networks Hasan Guclu.
Stopping computer viruses through dynamic immunization E. Shir, J.Goldenberg, Y. Shavitt, S. Solomon.
On Power-Law Relationships of the Internet Topology Michalis Faloutsos Petros Faloutsos Christos Faloutsos.
4. PREFERENTIAL ATTACHMENT The rich gets richer. Empirical evidences Many large networks are scale free The degree distribution has a power-law behavior.
Topology Generation Suat Mercan. 2 Outline Motivation Topology Characterization Levels of Topology Modeling Techniques Types of Topology Generators.
On the Spread of Viruses on the Internet Noam Berger Joint work with C. Borgs, J.T. Chayes and A. Saberi.
Scale Free and Small Worlds Networks: Studying A-Synchronous Discussion Groups Gilad Ravid Open University of Israel & Haifa University
Power law random graphs. Loose definition: distribution is power-law if Over some range of values for some exponent Examples  Degree distributions of.
Scale-free networks Péter Kómár Statistical physics seminar 07/10/2008.
The Barabási-Albert [BA] model (1999) ER Model Look at the distribution of degrees ER ModelWS Model actorspower grid www The probability of finding a highly.
1 Epidemic Spreading in Real Networks: an Eigenvalue Viewpoint Yang Wang Deepayan Chakrabarti Chenxi Wang Christos Faloutsos.
Web as Graph – Empirical Studies The Structure and Dynamics of Networks.
Physical Mechanism Underlying Opinion Spreading Jia Shao Advisor: H. Eugene Stanley J. Shao, S. Havlin, and H. E. Stanley, Phys. Rev. Lett. 103,
CS Lecture 6 Generative Graph Models Part II.
Physical Mechanism Underlying Opinion Spreading
The structure of the Internet. How are routers connected? Why should we care? –While communication protocols will work correctly on ANY topology –….they.
Agent-Organized Networks for Dynamic Team Formation Gaton, M.E. and desJardins, M., In Proceedings of AAMAS-2005, pp Seo, Young-Woo.
The structure of the Internet. The Internet as a graph Remember: the Internet is a collection of networks called autonomous systems (ASs) The Internet.

Error and Attack Tolerance of Complex Networks Albert, Jeong, Barabási (presented by Walfredo)
Internet Quarantine: Requirements for Containing Self-Propagating Code David Moore et. al. University of California, San Diego.
Economics of Malware: Epidemic Risk Model, Network Externalities and Incentives. Marc Lelarge (INRIA-ENS) WEIS, University College London, June 2009.
Peer-to-Peer and Social Networks Random Graphs. Random graphs E RDÖS -R ENYI MODEL One of several models … Presents a theory of how social webs are formed.
A Machine Learning-based Approach for Estimating Available Bandwidth Ling-Jyh Chen 1, Cheng-Fu Chou 2 and Bo-Chun Wang 2 1 Academia Sinica 2 National Taiwan.
Optimization Based Modeling of Social Network Yong-Yeol Ahn, Hawoong Jeong.
(Social) Networks Analysis III Prof. Dr. Daning Hu Department of Informatics University of Zurich Oct 16th, 2012.
1 Worm Modeling and Defense Cliff C. Zou, Don Towsley, Weibo Gong Univ. Massachusetts, Amherst.
1 Secure Cooperative MIMO Communications Under Active Compromised Nodes Liang Hong, McKenzie McNeal III, Wei Chen College of Engineering, Technology, and.
IEEE P2P, Aachen, Germany, September Ad-hoc Limited Scale-Free Models for Unstructured Peer-to-Peer Networks Hasan Guclu
Prof. A. Taleb-Bendiab, Talk: SOAS’07, Contact: Date: 12/09/2015, Slide: 1 Software Engineering Concerns in Observing Autonomic.
“The Geography of the Internet Infrastructure: A simulation approach based on the Barabasi-Albert model” Sandra Vinciguerra and Keon Frenken URU – Utrecht.
Senior Project Ideas: Blind Communication & Internet Measurements Mehmet H. Gunes.
Cache Management of Dynamic Source Routing for Fault Tolerance in Mobile Ad Hoc Networks.
Emergence of Scaling and Assortative Mixing by Altruism Li Ping The Hong Kong PolyU
Peer Pressure: Distributed Recovery in Gnutella Pedram Keyani Brian Larson Muthukumar Senthil Computer Science Department Stanford University.
Национальная процедура одобрения и регистрации проектов (программ) международной технической помощи (исключая представление информации об организации и.
Robustness of complex networks with the local protection strategy against cascading failures Jianwei Wang Adviser: Frank,Yeong-Sung Lin Present by Wayne.
Distance Estimation by Constructing The Virtual Ruler in Anisotropic Sensor Networks Yun Wang,Kai Li, Jie Wu Southeast University, Nanjing, China, Temple.
Decapitation of networks with and without weights and direction : The economics of iterated attack and defense Advisor : Professor Frank Y. S. Lin Presented.
1 On the Performance of Internet Worm Scanning Strategies Authors: Cliff C. Zou, Don Towsley, Weibo Gong Publication: Journal of Performance Evaluation,
Robustness and Structure of Complex Networks Shuai Shao Boston University, Physics Department A dvisor: Prof. Gene Stanley Co-advisor: Prof. Shlomo Havlin.
Class 9: Barabasi-Albert Model-Part I
1 Gossip-Based Ad Hoc Routing Zygmunt J. Haas, Joseph Halpern, LiLi Cornell University Presented By Charuka Silva.
Social Engineering © 2014 Project Lead The Way, Inc.Computer Science and Software Engineering.
Comparison of Tarry’s Algorithm and Awerbuch’s Algorithm CS 6/73201 Advanced Operating System Presentation by: Sanjitkumar Patel.
An Effective Method to Improve the Resistance to Frangibility in Scale-free Networks Kaihua Xu HuaZhong Normal University.
Mining information from social media
Project funded by the Future and Emerging Technologies arm of the IST Programme Search in Unstructured Networks Niloy Ganguly, Andreas Deutsch Center for.
Communications Range Analysis Simulation Set Up –Single Biological Threat placed in Soldier Field –Communication range varied from meters –Sensor.
An Improved Acquaintance Immunization Strategy for Complex Network.
1 Roie Melamed, Technion AT&T Labs Araneola: A Scalable Reliable Multicast System for Dynamic Wide Area Environments Roie Melamed, Idit Keidar Technion.
Hierarchical Organization in Complex Networks by Ravasz and Barabasi İlhan Kaya Boğaziçi University.
Incrementally Improving Lookup Latency in Distributed Hash Table Systems Hui Zhang 1, Ashish Goel 2, Ramesh Govindan 1 1 University of Southern California.
Scaling Properties of the Internet Graph Aditya Akella With Shuchi Chawla, Arvind Kannan and Srinivasan Seshan PODC 2003.
Algorithms and Computational Biology Lab, Department of Computer Science and & Information Engineering, National Taiwan University, Taiwan Network Biology.
Cmpe 588- Modeling of Internet Emergence of Scale-Free Network with Chaotic Units Pulin Gong, Cees van Leeuwen by Oya Ünlü Instructor: Haluk Bingöl.
Optimization of Blaster worms Performance Evaluation Laboratory s Tatehiro Kaiwa Supervised by Prof. Hiroshi Toyoizumi by Stochastic Modeling.
Epidemic Profiles and Defense of Scale-Free Networks L. Briesemeister, P. Lincoln, P. Porras Presented by Meltem Yıldırım CmpE
Internet Quarantine: Requirements for Containing Self-Propagating Code
Effects of External Links on the Synchronization of Community Networks
Brian Lafferty Virus on a Network.
Enhancing Attack Robustness of Scale-free Networks by Camouflage
Department of Computer Science University of York
Lecture 2: Complex Networks
Sampling Distributions (§ )
Presentation transcript:

Dynamics of Malicious Software in the Internet Tatehiro Kaiwa, University of Aizu. E-mail:m5081224@u-aizu.ac.jp I’d like to start my presentation. My name is Tatehiro Kaiwa. My thesis title is Dynamics of Malicious Software in the Internet. 1

Outline Random Network and Scale-free Network Observed Arrivals of E-mail Simulation Model of Worm Spread Dynamics Local Network Structure Inference Mathematical Model of Outbreak Hub Defense Strategy Conclusion This is outline First, I introduce relationship between random network and scale-free network. Next, I explain why I observed arrival time of e-mail. And, I explain my simulation model and results. Then, I explain Mathematical Model and I validate a method based on mathematical model. Finally, conclusion. 2

Two Model of Network Model of Network Random Network Degree Distribution: bell curve Scale-free Network Degree Distribution: power-law In this world, there are various networks – such as Highway, Social Network, the Internet and so on. We can categorize structure of such network into this two models. One is Random Network Model and Another One is Scale-free Network model. Comparing Degree Distribution of these models, we can see a difference like this.(図を指し示しながら) This figures show degree distribution of each. Degree Distribution of random network follows a bell curve whereas Degree Distribution of Scale-free Network follows power-law. ここから図を指しながら説明していく。 In random network, most nodes have the same number of links, and nodes with a very large number of links don’t exist. In contrast, power-law degree distribution predicts that most nodes have only a few links and that a few nodes have many links. In fact, many of networks in the world are categorized into Scale-free Network. 3

Scale-free and Preferential Attachment Why are there such networks. A keyword for understanding this is Preferential Attachment. Network grows daily. When a new node belongs to a network, the node does not connect to another one selected randomly. In real networks linking is never random. A subtle law of preferential attachment governs evolution of a network. Therefore, in such networks, some nodes attached preferentially become special nodes called hub nodes. As a newly added node attaches hub nodes preferentially, hub nodes have more and more links. Recently, structure of various network are Scale-free. For example, network of airline map, of epidemic and of connections in the Internet. Therefore, there is a probability that the e-mail network in the Internet is also Scale-free network. Scale-free Network is a network with power-law degree distribution. 4

Structure of E-mail Network *k: The number of links. Degree Distribution of an e-mail network. Reference: Holger Ebel, Lutz-Ingo Mielsch, and Stefan Bornholdt, “Scale-free topology of e-mail networks”, Physical Review E 66, 2002 This figure shows degree distribution of k. k is the number of links a node has. This data is observed in Kiel University. We can see that this is Scale-free because the graph decrease linear in double logarithmic plot. Therefore, I propose that we consider propagation of a worm on Scale-free Network. It is not clearly that how a worm propagates on Scale-free Network. A network in which we know topology is only a local network such this example. We don't have data to analyze the structure of network outside local network. Next, I will introduce a reason why we cannot analyze the structure of network outside local network. 5

Spoofed From-field The From-filed of an e-mail message a worm sends is varies and/or is spoofed. It is almost impossible to identify where a worm sends the e-mail and how many worms send observed e-mails. It is only arrival intervals that we can obtain a correct data from received e-mails. Because the From-field of an e-mail message a worm sends often varies and/or is spoofed, it is difficult to identify the PC a mass-mailing worm has infected and a correct data we can obtain from received e-mails is arrival time. Thus, a way by which we estimate a network structure is only arrival time. I thought that I could estimate a network structure from the arrival intervals. Therefore, I used observed data of arrival time of wild worms. 6

Observed Arrivals of E-mail There are log data* of the time on which each e-mail messages with a worm attached arrived at University of Aizu. Generally, we can observe arrival time of e-mail messages with a worm attached like this image. 図を指し示しながら。 When there are some infected PCs, they send e-mail messages with a worm attached to other address independently. E-mail messages we can observed are only the e-mails sent to an our address. Because the from-fields is spoofed, we don’t identify who send a observed e-mail. As I used arrival intervals of these, I use log data at University of Aizu. We can see the data from this URL. * http://web-int/labs/istc/ipc/Security/virus/index.html 7

Simulation Model of Worm Spread Dynamics I assume that the log data were observed such like this image. The PC object shows an e-mail address. Then, if two nodes shares address each other, the nodes are considered to have a link between them. I set the number of nodes in the community is 10000. I assume that the connectivity between the nodes in the community obeys power-law with Barabasi-Albert model. Brabasi-Albert model is one of a Scale-free Model. Each node has many addresses except links connecting to a node in the network. I suppose that the number of addresses excluding the addresses in the community is proportional to the number of links connected to nodes in the community. And a probability with which an infected node infects other neighbors is c. To fit a result of simulation to observed data, I estimate these parameter are nine times of links and 0.01 from results of many experiments. Then, in this simulation, each node does not be infected in 2 hours even though it receives an e-mail with a worm attached. This corresponds to intervals at which a receiver runs a mailer program. 8

Comparison between Simulation and Observed Data This figure shows result of our simulation compared with observed data of a worm, Netsky.P. Netsky.P is one of worms which infected many PCs in 2004. 9

Arrival Intervals of Simulation ii) iii) i) mk:115.619 ii) mk:92.15 iii) mk:61.95 Each graph shows a result on different observer. Time 0 is started time in simulation. Red line shows arrival intervals and Blue points shows time at which neighbors of each observer are infected. Arrival intervals in top left figure decrease slowly than others. This result shows the number of infected neighbors increases slowly and the number of links the neighbors have are large. Arrival intervals in bottom left figure decrease rapidly than others. This result shows the number of infected neighbors increases rapidly and the number of links the neighbors have are small. Arrival intervals in top right figure decrease slowly than bottom left one even though increase of the number of infected neighbors is similar. As mean of the number of links the neighbors have in top left one are larger than bottom left ones, we can see the result. From introducing results, we can estimate the structure of neighbors by comparing some observer. *mk : Mean of Number of links neighbors have. 10

Mathematical Model of Outbreak We assume that network size is enough and there is no infection loop in the network. S is the total number of infected nodes, and the infection started from a node selected randomly. We can obtain expectation of S from this equation. A parameter M is the number of links a randomly selected node has. However each one of links connects to a node which is liable to infect. Then a parameter Me is the number of links which a node which a randomly selected link connects to has. As hub nodes have extremely many links, a probability which a link selected randomly connects to a hub node with is high. So, In long-tailed distribution such as a power-law distribution, Expectation of M_e is greater than or equal expectation of M. The more the tail is long, the more expectation of M_e is large. From this equation, we can see that expectation of S diverges when (2 – Expectation of Me) equals 0. So we cannot prevent the outbreak. Decreasing E[M_e], we need to decrease Variance of M. Then, decreasing Variance of M, we need to protect hub nodes which have many links. I simulated an effect of this method. 11

Hub Defense Strategy (1) Difference of Number of immune hub nodes. These graph show an effect of hub defense strategy in changing number of immune hub nodes. Bottom one is result in early time. We can see that we obtain enough advantage in early time even though the number of immune hub nodes is much enough. *h = Number of immune hub nodes 12

Hub Defense Strategy (2) Comparison Between Hub Defense and Random Defense A purple line shows an effect when I make 3000 nodes selected randomly immune. We can see that we cannot obtain advantage by protecting random selected nodes. We can see that we obtain more advantage by protecting 1000 hub nodes better than by protecting 3000 selected randomly because outbreak is occurred in early time. r = Number of immune nodes selected randomly. h= Number of immune hub nodes. 13

Conclusion Observing arrival intervals, we can estimate damage of a worm and estimate a network structure around observer. We can confirm that hub defense strategy is an effective method in this network even though the number of immune hub nodes are not much enough. Conclusion. Observing arrival intervals, we can estimate damages by a worm and can estimate a network structure around observer. We can confirm that hub defense strategy is an effective method in this netowork even though the number of immune hub nodes are not much enough. 14

Thank you That’s all. Thank you for your listening. 15