WiCop: Engineering WiFi Temporal White-Space for Safe Operations of Wireless Body Area Networks in Medical Applications Yufei Wang, Qixin Wang, Zheng Zeng, Guanbo Zheng, Rong Zheng Dept. of Computing, The Hong Kong Polytechnic Univ. Dept. of Computer Science UIUC Dept. of Computer Science, Univ. of Houston

Table of Contents Demand Proposed Framework Evaluation Related work

Demand

Crisis of wired BAN Cause frequent falling off of medical sensors Limit the movement of patients Make medical unit untidy

Advantage of WBAN WBAN solves the problem of wired BAN Sensors unlikely fall off Patient feel more comfortable Medical units are more tidy

Medical WBAN Features Low duty cycle Low data rate ~500Kbps Typical sampling rate < 300Hz Wakeup on demand Low data rate ~500Kbps Low transmit power <1mW Disparate delay requirements ECG: <500ms Body temperature monitoring: several seconds Single-Hop centralized WBAN is the preferred architecture

WiFi Co-Channel Interference WiFi Co-channel inference is a major threat to WBAN ZigBee channels vs. 802.11b WiFi channels

WiFi Co-Channel Interference Power asymmetry Typical WiFi power ≈ 30mW Typical ZigBee (Bluetooth, IEEE 802.15.6) power ≤ 1mW MAC asymmetry Many WiFi device use Carrier Sense (CS) based Clear Channel Assessment (CCA). Such WiFi devices do not back off to ZigBee. Many ZigBee uses Energy Detection (ED) CCA to assess the channel. ZigBee backs off to WiFi.

The existing Solutions Operate WBAN over RF channels sufficiently away from the active WiFi RF channels limits the RF spectrum that WBANs can use Revise current WBAN or WiFi standards adding intelligent coexistence schemes Hard to use Commercially-Off-The-Shelf (COTS) Try to spatially separate WBANs from WiFi networks via careful configuration-time planning Often difficult as WiFi networks may not be under the same administration domain as WBANs Cause spurious outages in WBANs by unintended usage of mobile WiFi devices

The treat of WiFi to WBANs

The treat of WiFi to WBANs Monitor: Base station Polling period: 100ms Electrode: Client 250 samples/sec (4ms/sample) 25 samples/chunk (100ms/chunk) 3chunks/packet, i.e. each chunk is retransmitted 3 times (COTS≤4ms to send a packet)

The treat of WiFi to WBANs monitor: Base station polling period: 100ms electrode: Client 250 samples / sec 25 samples / chunk 3 chunks / packet, i.e., each chunk is retransmitted 3 times Failure: a chunk fails all of its retransmissions. a chunk fails all 𝑁𝑟𝑒 = 3 retransmissions.

Performance metrics 𝑴𝑻𝑻𝑭= 𝑻 𝒑𝒐𝒍𝒍𝒊𝒏𝒈 (𝟏−𝑷𝑹𝑹) 𝑵𝒓𝒆 Packet Reception Rate (PRR) Mean Time To Failure (MTTF) 𝑴𝑻𝑻𝑭= 𝑻 𝒑𝒐𝒍𝒍𝒊𝒏𝒈 (𝟏−𝑷𝑹𝑹) 𝑵𝒓𝒆

Proposed Framework

Proposed Framework “Engineer” temporal white-spaces between WiFi transmissions to allow WBAN transmissions Busy WiFi leaves no room for WBAN Goal: create temporal white-spaces in WiFi traffic for WBAN

Proposed Framework Policing: prohibit the transmissions of WiFi interferers in a well-controlled manner Shield WBAN transmissions in space and time The basic idea is to exploit the WiFi CCA mechanisms; sending WiFi compliant signals to refrain WiFi stations from transmitting.

WiCop Effectively controlling the temporal white-spaces (gaps) between consecutive WiFi transmissions Two mechanisms: Utilizing the carrier sensing mechanisms in WiFi Fake-PHY-Header DSSS Nulling

It is claims a fake WiFi packet with duration Fake-PHY-Header Temporal Scheme It is claims a fake WiFi packet with duration = WBAN active interval

Continuously repeated DSSS Preambles DSSS-Nulling Repeated DSSS preamble Continuously repeated DSSS Preambles

Band-rejection filtered DSSS-Nulling policing signal Spectrum illustration of interferer, policing and Zigbee signal

Evaluation

Evaluation The policing node implements the two policing mechanisms

Temporal white-spaces due to WiCop Without Policing With Policing (5ms temporal white-space/10ms)

Moderate Impact on WiFi traffic WBAN MTTF under different WiFi interference source end data rates

Related Work

WBAN and WiFi coexistence Huang [11] argued that the performance degradation of ZigBee in the WiFi interference is caused by two main reasons. Namely power asymmetry Carrier sense based CCA Hou [4] uses the duration field of the RTS MAC header to reserve time. Before broadcasting a beacon, a ZigBee base station first send an RTS packet to reserve a channel. Hou’s approach requires sending a whole packet to reserve the channel.

DoS attack to WLAN Thuente [29] studied several intelligent jamming methods with the requirement of Low power Low detection probability Including DIFS waiting jamming ACK corruption jamming Fake RTS jamming

Experimental Evaluation in Medical Environments Garudadri [32] applied Compressed Sensing to ECG. This approach uses the redundancy in periodic ECG trace, to mitigate distortion under high packet losses. This approach is orthogonal to WiCop and can be used in conjunction with WiCop to further improve the robustness of ECG monitoring.

Experiment layout

Conclusion WiCop significantly improves WBAN performance Controlled impact on WiFi DSSS-Nulling is more effective than Fake-PHY-Header in improving MTTF, mainly due to repeated transmissions of DSSS preamble Fake-PHY-Header incurs much less overhead than DSSS-Nulling