CSMA/CA characteristics of real-life IEEE 802.11 hardware - - PowerPoint PPT Presentation

.

.

CSMA/CA characteristics of real-life IEEE 802.11 hardware

Christopher Schramm

Chair for Network Architectures and Services Department for Computer Science Technische Universität München

April 3rd, 2013

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 1

.

Outline

1 Background 2 Observations 3 Goals 4 Motivation / Use 5 References 6 Timeline

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 2

.

Background

IEEE 802.11 uses CSMA/CA to avoid collisions CSMA/CA requires a sender to wait for a random backoff time between detecting a free medium and sending Distributed Coordination Function (DCF)

Busy medium Backoff slots Next frame Defer access Slot time Backoff interval Contention window DIFS SIFS

BackoffTime = Random() × aSlotTime Random() = Pseudorandom integer taken from contention window, which increases with failed transmissions

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 3

.

Observations

Hardware / driver vendors want their design to... ...be simple and cheap. ...possibly perform better than other devices. During his analysis of monitoring and injection capabilities, Stephan found these injection delays:

100 200 300 400 500 600 700 800 900 1,000 5 10 15 20 25 30 Echo request SEQ Injection delay [ms]

AR9380 AR9280 BCM43224 RT2870

Broadcom chip has no randomness at all Ralink is at least not statistically independent

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 4

.

Is this fair?

Is this fair? ⇒ No!

AR9280 AR9380 BCM43224 RT2870 BCM43224 AR9280 RT2870 AR9280 RT2870 BCM43224 5 10 15 20 25 30 35 40 45 50 55 Achievable rate [Mbit/s]

T X A → B R X A → B T X A ← B R X A ← B T X A A ↔ B T X B A ↔ B R X B A ↔ B

Neither is it clever. IEEE 802.11 9.3.3: It is important that designers recognize the need for statistical independence among the random number streams among STAs.

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 5

.

Goals

We want to further analyze those characteristics for real-life hardware in several steps: Monitor Mode Remote observation Managed Mode (needs remote observation) Different drivers / operating systems (Are observed anomalies related to specific driver / OS / hardware?) An overall ”bonus” goal would be to implement a sniffer that detects hardware, OS and / or driver by looking for specific characteristics in

• bserved traffic.

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 6

.

Motivation / Use

Monitor Mode characteristics need to be known when using frame injection, in our case for building of mesh networks Managed Mode characteristics may be of interest when choosing hardware for large network setups Detection may be of interest in security and administrative contexts to find vulnerable, flawed or deprecated hardware

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 7

.

References

Primary reference: IEEE 802.11

• T. Ho and K. Chen: Performance Analysis of IEEE 802.11

CSMA/CA

• G. Bianchi: Performance Analysis of the IEEE 802.11 Distributed

Coordination Function

• E. Ziouva and T. Antonakopoulos: CSMA/CA Performance

under High Traffic Conditions: Throughput and Delay Analysis

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 8

.

Timeline

Official schedule: March 15th - September 15th Monitor Mode: March and April Remote observation: May Managed Mode: June Different drivers / OSs: July Detection by characteristics: August Last two weeks dedicated to finishing the thesis

. Christopher Schramm: CSMA/CA characteristics of real-life IEEE 802.11 hardware 9