MAC LAYER MISBEHAVIOR EFFECTIVENESS AND COLLECTIVE AGGRESSIVE REACTION APPROACH. Department of Electrical Engineering and Computer Science

Transcription

1 MAC LAYER MISBEHAVIOR EFFECTIVENESS AND COLLECTIVE AGGRESSIVE REACTION APPROACH Vamshikrishna Reddy Giri Neeraj Jaggi Department of Electrical Engineering and Computer Science

2 Outline Introduction MAC layer Misbehavior - An overview Existing Misbehavior detection and reaction schemes Contributions of this paper Classification of various MAC layer Misbehaviors Related to Binary Exponential Backoff in wireless networks Measuring the effectiveness of various misbehaviors Effectiveness Metric formulation Measuring effectiveness using Opnet Simulations Proposed aggressive reaction approach Reduces Misbehavior effectiveness Provides fairness in the network Conclusion and Future work

3 MAC layer Misbehavior - An Overview Types of Misbehaving nodes at MAC layer Selfish nodes Mild cheating with respect to protocol Motive: In order to get higher throughput (bandwidth) Malicious nodes Disrupting network by Jamming etc Motive: In order to disrupt or collapse the network Achieving Misbehavior objectives by deviating from standard protocol behavior Modify Existing protocol ( BEB) Choosing smaller Backoff values than expected Other parameters (BEB Binary Exponential Backoff) SIFS, DIFS (if node does not wait for DIFS, it could get channel access faster)

4 Related Research Detection Schemes Sequential Probability Ratio Test [S. Radosavac, J. S. Baras, and I. Koutsopoulos, 2005] Based on observed Backoff intervals chosen by other nodes Access point based adaptive mechanism [M. Raya, I. Aad, J.-P. Hubaux, and A. E. Fawal, 2006] Reaction Schemes Receiver assigned Backoff [P. Kyasanur, and Nitin H. Vaidya, 2005] Concept of receiver assigning Backoff values to the sender Identify and isolate the Misbehaving node [Guang, Assi, and Ye 2007] Effects higher layer performance (e.g. connectivity, routing)

5 Contributions of this Paper Motivation Classification of various types of Misbehaviors Studying common characteristics and differences Exploring methods to penalize Misbehavior (other than Isolation) While guaranteeing fairness Contributions Classify and study different types of Misbehavior Outline common characteristics Measure Misbehaving node s incentive Define Effectiveness measure Aggressive reaction Approach Provides fairness Provides disincentive

6 Binary Exponential Backoff (BEB) A node which has data to transmit Chooses b uniformly from interval [0... CW 1] Channel is sensed idle Node waits for b time slots before accessing the channel. Channel is sensed busy Node freezes its Backoff until the channel is sensed idle again and continues counting down thereafter. Initial CW = CW min Successful transmission Resets CW to CW min Unsuccessful transmission Sets CW to min{2 * CW, CW max } CW min Minimum contention window size CW max - Maximum contention window size CW min = 32 CW max = CW - Current contention window size b - Backoff value of the node

7 Classifying MAC layer Misbehaviors Based upon modifications to BEB Algorithm Five types of misbehaviors considered Alpha, Beta, Fixed Contention window, Maximum contention Window, Deterministic Back off Vary the level (or aggressiveness) using parameters (α, β etc) Compare them in terms of Misbehavior effectiveness Hybrid Misbehaviors Combining two or more of the above Is there any additional incentive in using a hybrid instead

8 Alpha(α) Misbehavior Instead of choosing the Backoff b uniformly at random from the interval [0... CW 1] Selfish node chooses b uniformly at random from the interval [0... α(cw 1)] α Є [0, 1] Effect Node ends up choosing a smaller Backoff interval Increases its chances of accessing the channel next Smaller value of α leads to more aggressive misbehavior

9 Beta(β) Misbehavior Upon unsuccessful transmission, the selfish node sets its CW (instead of setting CW = min{2*cw, CW max }) as CW = max{cw min, min{β CW, CW max }}, β Є [0, 2] CW min = min{32, β 32} (useful when β < 1) CW cheating rather than Backoff cheating (as in α-misbehavior) Effect Similar to α Misbehaviors with smaller β are more aggressive

10 Fixed contention window (CW fix ) Misbehavior Selfish node sets its contention window to a small Fixed size CW fix Cheating with respect to CW The value CW fix is chosen both in case of successful and unsuccessful transmissions Chooses its Backoff interval Uniformly at random from the interval [0... CW fix ] Effect Aggressiveness of Misbehavior increases with decrease in CW fix Lower CW fix will yield lower Backoff values, increasing the channel access probability of Misbehaving node

11 Fixed Maximum Contention Window (CW max ) Misbehavior Choice of CW max For genuine nodes CW max equals However, the selfish node sets its maximum contention window to be a value smaller than Also the node sets CW min = min{32, CW max } No Backoff or CW cheating Effect More effective only when contention (traffic) is high Under normal scenarios CW max is seldom reached Ineffective unless CW max is quite low ( 64) (as observed)

12 Deterministic Backoff (db) Misbehavior Changes to BEB Instead of choosing the Backoff b uniformly at random from the interval [0... CW 1], Selfish node chooses a constant Backoff interval b which is, Independent of the current contention window size. Node cheats with respect to Backoff (CW not used at all). Effect Lower the values of constant Backoff chosen, the more aggressive is the Misbehavior

13 Effectiveness Measure Definition -: The percentage increase in the throughput of the selfish node gained via Misbehaving, compared to the scenario when all the nodes are genuine. Effectiveness t m - throughput of the node x, x being Misbehaving node and rest are genuine nodes t g - throughput of the node x, when all nodes are genuine

14 Simulation Scenario Network scenario Wireless LAN with data rate 2 Mbps Opnet Simulator 10 nodes (9 senders and single receiver) One sender Misbehaves if configured Traffic Scenario Exponential packet arrivals Packet size 512 bytes Profiles High load (100 packets per second ) Medium load 77 packets per second) Low load (25 packets per second)

15 Analyzing Effectiveness (α, β) Misbehavior Low load LAN traffic is below capacity Constant throughput No additional gain from Misbehavior Medium and High Load Non linear increase in Throughput with decrease in α, β Saturates - when Misbehaving node gets all its data across

16 Analyzing Effectiveness (db) Misbehavior Characteristics are similar to that of α and β Misbehaviors Non-linear increase, Saturation Average Backoff value of a genuine node in steady-state 22 If Misbehaving node employs Constant Backoff > 22 No gain achieved, resulting in negative effectiveness

17 Analyzing Effectiveness (CW fix, CW max ) Misbehavior Negative Effectiveness Average CW at steadystate for a genuine node 50 Hence negative effectiveness for CW > 50 Genuine nodes Generally do not contend with CW max = 1024 Aggressive only at lower values of CW max

18 Analyzing effectiveness (Hybrid) Misbehavior Two hybrid Misbehaviors With α and CW max = 64 With α and β = 1.5 Characteristics are similar to individual Misbehaviors However at a given α Effectiveness in hybrids is greater than that using only α Misbehavior

19 Proposed Aggressive Reaction Approach Objective of a reaction scheme Selfish node s throughput should become less than what it would have been in the absence of any Misbehavior Therefore no incentive for Misbehavior (negative incentive!) Proposed Approach to realize objective Genuine nodes estimate level of Misbehavior in the network Replicate the Misbehavior themselves (Collective Misbehavior) Fairness achieved at the cost of overall LAN throughput degradation Assumption Genuine nodes can detect the level of Misbehavior in the network

20 Effect on Overall LAN Throughput All nodes are misbehaving with α Misbehavior With lower values of α (Aggressive Misbehavior) Overall LAN Throughput degrades However decrease in the LAN throughput also includes a decrease in Throughput of the Misbehaving node

21 Effect on Fairness Jain s Fairness Index (f) Value 1 corresponds to maximum possible fairness f is computed for Collective α Misbehavior for various α values Value achieved Bar graph represents the throughput share of all nodes employing CW fix = 2 Misbehaving node s Throughput equal to that of genuine nodes Reaction Scheme achieves fairness

22 Conclusion and future work Classified various MAC layer Misbehaviors Studied their impact on throughput and fairness Effectiveness increases non-linearly with increase in level of Misbehavior Effectiveness saturates (when all the data gets through) Higher traffic load changes the point of saturation Employing Hybrid Misbehavior provides no additional advantage α and CW fix Most effective Misbehaviors Proposed Collective Aggressive reaction response Ensures fairness in the network Provides disincentive to cheating Work in Progress Estimating the level of Misbehavior in the network