PHY layer access misbehavior in WLAN networks Master thesis presentation Radio Communication Systems, KTH Probir Khaskel Advisor: Olav Queseth & Examiner: Prof. Jens Zander
2 Outline Problem definition System model Link adaptation Single cell system Multi-cell system Game theoretical analysis Conclusion Future work Ideal system performance System response in CCA level modification Airport environment Infrastructure based network Time driven system in MATLAB 1 time slot = 1 OFDM symbol (4μ sec ) Players attitude Dynamic strategies for stable system
3 Problem Definition Is there any incentive to modify CCA level defined by the standard from the user’s point of view ? Greedy users in a single network Modification in the PHY layer Frequent channel access by modifying CCA threshold level ?? Does this modification have impact on the overall system performance (in terms of throughput) ? Greedy users in a single network Modification in the PHY layer Frequent channel access by modifying CCA threshold level
4 Greedy channel access in the PHY layer Threshold (dBm) Time ~ Channel is busy -62 Channel is Idle RF signal strength Problem background ↑
5 System model Propagation model Traffic model and packet System deployment Transmit power and noise Capture model Channel interference Traffic model and packet Asynchronous and identical traffic from the upper layer Fixed packet size (MPDU-256 bytes) System deployment Single cell system with various number of nodes Multicell system with 16 cells, re-use factor 4, 3 nodes in each cell Transmit power and noise Fixed maximum power of +23 dBm Noise level -95 dBm Capture model i ij jjrx ii NdP dP )( )(,, Channel interference Co-channel interference Adjacent channel interference is ignored Propagation model
6 IEEE a PHY and link adaptation 8 PHY modes with data rates ranging from 6 to 54 Mbps Link Adaptation for data transmission is realized as MPDU-based fast link adaptation, placed closer to the air-interface based on the estimated C/I at the receiver Link Adaptation for ACK transmission is realized as receiver adopts the same PHY mode as the corresponding received data packet 8 PHY modes with data rates ranging from 6 to 54 Mbps Link Adaptation for data transmission is realized as MPDU-based fast link adaptation, placed closer to the air-interface based on the estimated C/I at the receiver Link Adaptation for ACK transmission is realized as receiver adopts the same PHY mode as the corresponding received data packet C/I (dB) 54 Mbps 48 Mbps 36 Mbps 24 Mbps 18 Mbps 12 Mbps 9 Mbps 6 Mbps Data Rate 3/464-QAM8 2/364-QAM7 3/416-QAM6 1/216-QAM5 3/4QPSK4 1/2QPSK3 3/4BPSK2 1/2BPSK1 Code RateModulationMode
7 Single cell system
load vs. throughput avg. offered load per station [Mbps] avg. throughput [Mbps] 6 STA 11 STA 16 STA 21 STA load vs. delay avg. offered load per station [Mbps] avg. delay [ms] For higher number of users, system gets saturated earlier Single cell: Ideal system performance System with higher number of users is less capable to support delay bounded QoS with increasing offered load More number of users in the system, more throughput drops down from system capacity to saturation level Number of collision also gets saturated in system saturation load vs. collision avg. offered load per station [Mbps] avg. no. of collision
9 The less number of users in the system, the higher reachable CCA level Single cell: Saturation analysis
10 Single cell: Greedy vs. timid users
11 load vs. throughput avg. offered load per station [Mbps] avg. throughput [Mbps] G2: -52 dBm G1: -50 dBm Ideal system avg. system, (D|D) System throughput decreases than that of the ideal system Number of collision increases around eight times, however, delay performance does not deteriorate compare to throughput and collision load vs. delay avg. offered load per station [Mbps] avg. delay [ms] G2: -52 dBm G1: -50 dBm Ideal system avg. system, (D|D) load vs. collision avg. offered load per station [Mbps] avg. collision G2: -52 dBm G1: -50 dBm Ideal system avg. system, (D|D) Single cell: Greedy vs. greedy users
12 Current Nash Equilibrium is Pareto inefficient (S|S) would be Pareto efficient NE System performance is better when users follow the standard protocol Current Nash Equilibrium is Pareto inefficient (S|S) would be Pareto efficient NE Single cell: single stage game G1↓ G2→SD S(0.719, 0.719)(0.499, 0.761) D(0.857, 0.466)(0.612, 0.521)
13 Single cell: multi stage game Users prefer to defect if δ<0.563, meaning that they are more likely to defect Dynamic strategies: TFT, GRIM ensure a stable system Discount factor, δ (delay bounded application dependent) Anticipated payoff in stage t to player i In an infinite game, payoff is computed as
14 Multi-cell system
15 System throughput out-perform the standard by cooperative modification of the CCA threshold Collision reduces around three and a half times compare to the ideal system Multi-cell: saturation analysis
16 G1↓ G2→CD C(0.665, 0.665)(0.327, 0.438) D(0.727, 0.398)(0.366, 0.366) SSG: Current Nash Equilibrium is Pareto efficient MSG: TFT, GRIM dynamic strategies ensure a stable system Users prefer to defect if δ<0.171, meaning that they are more likely to cooperate Multi-cell: game analysis
17 Conclusion Single cell system It’s possible to achieve higher throughput by modifying the CCA level Any modification results in deterioration of the system performance Multi-cell System Adaptive modification of the CCA level gives a noticeable system improvement A small group gains by further modification, the overall system performance deteriorates Users are more likely to cooperate Operators might be interested to have a control on the CCA level modification based on the network condition and update the users to adjust in a regular fashion
18 Future work Part of the received data of a collided packet could be recoverable by smart decoding algorithm, which in tern could increase the system throughput by avoiding to retransmit the whole packet Transmit Power Control (TPC) could increase system capacity by minimizing co-channel interference In general, any misbehaving activities can be detected by collision counter. However, pinpointing a misbehaving user is a crucial task Players’ assessment of others’ strategy by observed throughput might be a pitfall for system stability
19 Question&Comments!
δ → 0 1 Higher preference of future payoff, e.g. voice telephony Lower preference of future payoff, e.g. best effort type application Discount factor, δ Extra slides
21 Hidden Terminal Problem Hidden terminal Station/Node Access Point Extra slides
22 Unlicensed frequency bands UPCS-Unlicensed Personal Communication Services [1.9GHz] ISM-Industry, Science and Medicine [ GHz] UNII-Unlicensed National Information Infrastructure [ GHz] Why Unlicensed ? Promotes efficient spectrum sharing Further experimentation and innovation Mobility of wireless applications since no license needed in new location Introduction Extra slides