Advisor: Professor Frank Y.S. Lin Present by Tim Q.T. Chen

Transcription

1 Advisor: Professor Frank Y.S. Lin Present by Tim Q.T. Chen 1

2 Introduction Game Theory Attack Graph A Game Theoretic Method for Decision and Analysis of the Optimal Active Defense Strategy Optimal Network Security Strengthening Using Attack- Defense Game Model Future Work 2

3 3

4 The interactive behavior between the attacker and the defender is similar to information warfare. Both attacker and the defender may have several available strategies. Game theory can provide us with a mathematical framework for analysis and modeling network security problems, and it can support to find out the optimal strategy for all party. 4

5 5

6 Game theory describes multi-person decision scenarios as games where each player chooses actions which result in the best possible rewards for self, while anticipating the rational actions from other players.[2] Game theory can be used in the fields of economics, biology, political science computer science, and philosophy. 6

7 Game: A description of the strategic interaction between players. Player :A strategic decision maker within the context of the game (attacker and defender) Strategy :Plan of action within the game that a given player can take during game play. 7

8 Payoff : The payout a player receives from arriving at a particular outcome. Equilibrium : The point in a game where both players have made their decisions and an outcome is reached. 8

9 Cooperative or non-cooperative Zero-sum and non-zero-sum Simultaneous and sequential One-player and many-player Perfect information and imperfect information Complete information and incomplete information 9

10 Static Game : A static game is one in which all players make decisions simultaneously. (one-shot game) Dynamic Game: A game with more than one stages in each of which the players can consider their action. (sequential) 10

11 Normal Form Player 1 No. of strategy =2 Player 1 chooses top Player 1 chooses bottom Player 2 chooses left Player 2 chooses right 4, 3 1, 1 0, 0 3, 4 Player 2 No. of strategy =2 A pure strategy provides a complete definition of how a player will play a game. A mixed strategy is an assignment of a probability to each pure strategy. Payoff 11

12 Cooperative or non-cooperative Zero-sum and non-zero-sum Simultaneous and sequential One-player and many-player Perfect information and imperfect information Complete information and incomplete information 12

13 Extensive Form 2 player Two strategies of player 2 4, 3 0, 0-1, -1 3, 4 Two strategies of player 1 payoff 13

14 Cooperative or non-cooperative Zero-sum and non-zero-sum Simultaneous and sequential One-player and many-player Perfect information and imperfect information Complete information and incomplete information 14

15 Game theory Non-Cooperative game Cooperative game Static Dynamic 15

16 Static Complete and imperfect game Incomplete and imperfect game Zero sum Non-zero sum Zero sum Non-zero sum 16

17 Dynamic Complete and imperfect game Incomplete and perfect game Complete and imperfect game Incomplete and imperfect game Zero sum Nonzero sum Zero sum Nonzero sum Zero sum Nonzero sum Zero sum Nonzero sum 17

18 18

19 Knowing that a network system is not perfectly secure, the defenders can use attack graphs to analyze security vulnerabilities in enterprise networks. Ex.Nessus [4][5] From [3] 19

20 The number of vulnerability of ip1 =2 Objective The number of vulnerability of ip2 =1 The topology of network The attack graph The number of attack path =3 20

21 2007 Wei Jiang Zhi-hong Tian Hong-li Zhang Xin-fang Song 21

22 Introduction The ADG Model Formalization Attack-Defense Game Strategies for the Two Players Cost, Reward and Payoff Matrix OADSD Algorithm Numerical Analysis Conclusion and Future Work 22

23 This paper presents a game-theoretic method for analyzing the active defense of computer networks. Attack-defense game(adg) model two-player (attacker and defender) non-cooperative zero-sum 23

24 An optimal active defense strategy decision (OADSD) algorithm is developed using ADG cost-sensitive model ADG Costsensitive model OADSD Game Payoff Optimal Strategy of Defender 24

25 Optimal defense strategies with minimizing costs (payoff value) are used to defend the attack and harden the network in advance. 25

26 An ADG is a two-player, zero-sum, noncooperative, finite game G=({attacker, defender}, { Sa, Sd}, {Ua, Ud}) Two Player Strategy Space Sa is an attack strategy (the number of strategy is m) 26

27 For attacker: node 1:initial state node 6:success state Strategy set of attacker=3 27

28 For defender: The defender s strategies for attacker s strategies get from known defense strategy set. 28

29 TDcost = DOcost(d) + RDcost(a) + DN cost(d) total defense cost defense operation cost residual damage cost defense negative cost the amount of resources needed to defense an attack (type). L1 cost 1~5 L2 cost 10 L3 cost 100 Defense strategy cannot defend the attack that may occur in the future result in some attack damage. DNcost represents an effect of a defense action on a target system 29

30 1. An outcome in a matrix ADG is called a saddle point or pure min-max equilibrium if the entry at that outcome is the minimum in its row, and the maximum of its column. If an ADG matrix has a saddle point, both players should play a strategy which the saddle point stipulates. (Pure strategy) 30

31 If there is no saddle point in the matrix ADG, we can solve it by linear program. (mixed strategy) 31

32 P is the attacker has a probability distribution over the set of attack path. q is the defender has a probability distribution over the Set of defender strategy. i is any probability distribution over the attack path. The object finds the probability distribution of q (defender) 32

33 Payoff value Linear program 33

34 34

35 35

36 Environment Nessus 弱點偵測軟體 DB 36

37 37

38 38

39 Sa={,,,, } 39

40 The strategy of the defender: 40

41 The payoff matrix: The strategies of defender The strategies of attacker =IP Blocking is Optimal Strategy DOcost+RDcost+DNCost =10+Ɛ*DCost(a)+0 =10+(0*50+1*100)+0 =110 41

42 We have presented a game-theoretic method for analyzing optimal active defense strategy decision in networks security. The game is: two-player (attacker and defender) non-cooperative zero-sum 42

43 1. We would investigate attack taxonomy because it is essential in producing meaningful cost metrics. 2. We would investigate cost factors and quantified analysis of defense and attack. 43

44 2009 Wei Jiang Zhi-hong Tian Hong-li Zhang Xin-fang Song 44

45 Introduction Model Definition and Formalization Attack Taxonomy and Defense Strategy Attack and Defense Cost Metrics Defense Graph Model Attack-Defense Game Model Optimal Network Strengthening Strategy Selection Algorithm Experiment Results Conclusion and Future Work 45

46 46

47 TDcost = DOcost(d) + RDcost(a) + DN cost(d) total defense cost defense operation cost residual damage cost negative damage the amount of resources needed to defense an attack (type). L1 cost 1~5 L2 cost 10 L3 cost 100 Defense strategy cannot defend the attack that may occur in the future result in some attack damage. DNcost represents an effect of a defense action on a target system 47

48 The node A:initial state The node F:final state The number of Atomic attack The defense strategy of attack path The strategy of the attacker : 48

49 Environment The attack path: 1. 先攻 web server, 才能攻 DB 2. 先攻 FTP, 才能攻 DB 3. 先攻 SMTP, 才能攻 DB 49

50 使用 Nessus 弱點偵測軟體 Objective 50

51 51

52 The strategy number of attacker=4 52

53 The payoff matrix : The strategy of defender The strategy of attacker 53

54 1.The cost model that we adopt in this paper is somewhat simplistic. 2. We would investigate the automatically generation of defense graphs. 3.How to demonstrate the cost parameters in our ADG model will affect the expected attacker behaviors and selection of optimal strengthening strategies. 54

55 55

56 1.Using different ways to calculate the payoff? [6] 56

57 2. Toward more realistic network defense in depth 57

58 3. K-round 58

59 4.Zero sum or non-zero sum? 59

60 60

61 [1] A Game Theoretic Approach to Decision and Analysis in Strategies of Attack and Defense. [2] A Survey of Game Theory as Applied to Network Security [3] Scenario Graphs and Attack Graphs [4] A Scalable Approach to Attack Graph Generation [5] Two Formal Analyses of Attack Graphs [6] Applying Attack Graphs to Network Security Metric 61