Device interaction at the Physical layer What the RF-channel can - - PowerPoint PPT Presentation

Device interaction at the Physical layer

What the RF-channel can tell us

Stephan Sigg

Milton Keynes, 05.07.2012

Introduction Recognition Calculation Security Conclusion

Introduction

Stephan Sigg | Physical layer device interaction | 2

Introduction Recognition Calculation Security Conclusion

Introduction

IoT environments feature communication, sensing and actuating

These devices contain a great diversity of sensing hardware and differ in their design, power consumption, size or purpose.

Stephan Sigg | Physical layer device interaction | 2

Introduction Recognition Calculation Security Conclusion

Introduction

IoT environments feature communication, sensing and actuating

These devices contain a great diversity of sensing hardware and differ in their design, power consumption, size or purpose. They will all have a single interface in common:The RF-interface

Stephan Sigg | Physical layer device interaction | 2

Introduction Recognition Calculation Security Conclusion

Aspects of the mobile radio channel

Stephan Sigg | Physical layer device interaction | 3

Introduction Recognition Calculation Security Conclusion

Aspects of the mobile radio channel

Stephan Sigg | Physical layer device interaction | 4

Introduction Recognition Calculation Security Conclusion

Introduction

How to utilise this common interface for Ubiquitous applications?

Environmental sensing Calculation of mathematical functions Secure communication based on RF

Stephan Sigg | Physical layer device interaction | 5

Introduction Recognition Calculation Security Conclusion

Outline

Introduction RF-based activity recognition Calculation during transmission on the wireless channel Security from environmental stimuli Conclusion

Stephan Sigg | Physical layer device interaction | 6

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

In the IoT wireless interfaces will be virtually everywhere

Can we re-use this hardware to gain additional value? The RF-channel is a ubiquitous source of environmental information Multi-path propagation Scattering Reflection Blocking of signal paths

Stephan Sigg | Physical layer device interaction | 7

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Device free localisation (DFL)a

• aM. Youssef et al., Challenges: Device-free passive localisation for wireless environments,

MobiCom2007

Localisation from RF-data RSSI-based passive localisation

Monitoring of breathinga

• aN. Patwari et al., Spatial models for human motion-induced signal strength variance on

static links, IEEE Transactions on Information Forensics and Security, vol. 6, no. 3, pp. 2011

Two-way RSSI measurements Accuracy: 0.1 to 0.4 beats

Stephan Sigg | Physical layer device interaction | 8

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Sensewaves Video

Stephan Sigg | Physical layer device interaction | 9

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Can we use RF-channel information for activity detection?1 1 USRP transmitter 1 USRP receiver 900 Mhz constant signal 2 individuals 5 activities 15 locations

1Stephan Sigg, Markus Scholz, Yusheng Ji, Michael Beigl, Active and Passive sensing of activities from amplitude-based RF-features in device-free recognition systems, (Submitted to IEEE Transaction on Mobile Computing, 01.2012) Stephan Sigg | Physical layer device interaction | 10

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Activities and location detected with high accuracy Location-accuracy about 1-2 meters Lying detected worst Transmitter and receiver under complete control

Classified Truth crawling walking empty lying standing crA .986 .013 .001 waA .019 .975 .006 emA .911 .031 .059 lyA .01 .2 .704 .086 stA .215 .005 .78 crB .998 .002 waB .016 .979 .004 .002 emB .937 .004 .058 lyB .003 .191 .719 .086 stB .007 .001 .252 .005 .736 crC .998 .002 waC .009 .986 .005 emC .002 .933 .005 .061 lyC .001 .156 .646 .197 stC .001 .008 .17 .012 .808 crD .933 .065 .001 .001 waD .034 .964 .001 .001 emD .021 .889 .004 .085 lyD .191 .725 .084 stD .015 .143 .012 .829 crE .991 .001 .008 waE .997 .003 .001 emE .937 .005 .059 lyE .001 .238 .676 .086 stE .001 .005 .157 .016 .82 Stephan Sigg | Physical layer device interaction | 11

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Potential and challenges

The advancing IoT adds to the penetration by RF-capable devices

This fosters the evolution of smart environments to support activity recognition and awareness.

Stephan Sigg | Physical layer device interaction | 12

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Potential and challenges

The advancing IoT adds to the penetration by RF-capable devices

This fosters the evolution of smart environments to support activity recognition and awareness. Sustainable: Transceiver might suffice for common classification tasks; Re-use of RF-interface (already part of the design) No additional cost for further sensors

Stephan Sigg | Physical layer device interaction | 12

Introduction Recognition Calculation Security Conclusion

RF-based activity recognition

Potential and challenges

The advancing IoT adds to the penetration by RF-capable devices

This fosters the evolution of smart environments to support activity recognition and awareness. Sustainable: Transceiver might suffice for common classification tasks; Re-use of RF-interface (already part of the design) No additional cost for further sensors Potential: Improve features to cover less static scenarios Identify features that do not require training Tolerate static environmental changes Can we detect at the time the device is carried/moved

Stephan Sigg | Physical layer device interaction | 12

Introduction Recognition Calculation Security Conclusion

Outline

Introduction RF-based activity recognition Calculation during transmission on the wireless channel Security from environmental stimuli Conclusion

Stephan Sigg | Physical layer device interaction | 13

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

IoT devices will frequently draw energy from the environment

Devices will be sharply restricted in their computational resources This contradicts traditional WSN paradigms

Stephan Sigg | Physical layer device interaction | 14

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Motivation: Computation during transmissiona

• aA. Giridhar and P. Kumar, Toward a theory of in-network computation in wireless sensor

networks, IEEE Comm. Mag., vol. 44, no 4, pp. 98-107, april 2006

• Max. rate to compute & communicate functions

Mention: Collisions might contain information

Calculation of by means of post- and pre-processinga

• aM. Goldenbaum, S. Stanczak, and M. Kaliszan, On function computation via wireless

sensor multiple-access channels, IEEE Wireless Communications and Networking Conf., 2009

Requires accurate channel state information Requires identical absolute transmit power

Stephan Sigg | Physical layer device interaction | 15

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Utilising Poisson-distributed burst-sequences

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

t K burst

superimposed received burst sequence transmit burst sequences time

Addition, subtraction, division and multiplication at the time of wireless data transmission via Poisson-distributed burst-sequences Adding Poisson processes i with mean µi will result in a Poisson process with mean n

i=1 µi.

Stephan Sigg | Physical layer device interaction | 16

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Errors for calculating during transmission on the wireless channel

t = 106; κ = 103 10 nodes 20 nodes 30 nodes 40 nodes 50 nodes mean err .0322 .0466 .0609 .051 .0719 std-dev. .0232 .0368 .0536 .0336 .0438 max Ni 9 14 18.5 26 31 median T 2653.5 5161.5 7393 101816 124179 t = 107; κ = 103 10 nodes 20 nodes 30 nodes 40 nodes 50 nodes mean err .0049 .0176 .0402 .0475 .0781 std-dev. .0062 .0127 .0233 .0292 .0405 max Ni 12 18 23 27 31 median T 25708.5 52617.5 78502 101381 114348 t = 107; κ = 102 10 nodes 20 nodes 30 nodes 40 nodes 50 nodes mean err .0190 .1337 .2619 .4903 .6597 std-dev. .0107 .0358 .0591 .0708 .1129 max Ni 9.5 16 19 24 27 median T 24165 50037 71686.5 96829 114383

Stephan Sigg | Physical layer device interaction | 17

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Case study to compare the calculation accuracy Utilise data from the Intel Berkeley laboratory network (here: temperature)2 Transmission of data by simple sensor nodes

2http://db.csail.mit.edu/labdata/labdata.html Stephan Sigg | Physical layer device interaction | 18

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Offline Online

Offline Online

Stephan Sigg | Physical layer device interaction | 19

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Potential and challenges

Humans might become the minority to generate computational load

The sharply restricted resources of the majority IoT devices hosting this load calls for new ways of processing data.

Stephan Sigg | Physical layer device interaction | 20

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Potential and challenges

Humans might become the minority to generate computational load

The sharply restricted resources of the majority IoT devices hosting this load calls for new ways of processing data. Sustainable: IoT device’s design can maximise cost efficiency No need for high processing power or big storage Towards parasitic, passive nodes

Stephan Sigg | Physical layer device interaction | 20

Introduction Recognition Calculation Security Conclusion

Calculation during transmission on the wireless channel

Potential and challenges

Humans might become the minority to generate computational load

The sharply restricted resources of the majority IoT devices hosting this load calls for new ways of processing data. Sustainable: IoT device’s design can maximise cost efficiency No need for high processing power or big storage Towards parasitic, passive nodes Potential: Paradigm shift in distributed computing Not energy but processing power is the limit Trade processing load for communication load Explore further probability distributions to support

• ther operations

Stephan Sigg | Physical layer device interaction | 20

Introduction Recognition Calculation Security Conclusion

Outline

Introduction RF-based activity recognition Calculation during transmission on the wireless channel Security from environmental stimuli Conclusion

Stephan Sigg | Physical layer device interaction | 21

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Stephan Sigg | Physical layer device interaction | 22

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

IoT increases opportunities for attacks on personal data

Spontaneous communication among IoT devices should be secured automatically Utilise context to automatically secure communication Unobtrusively among not acquainted devices More natural perception of security

Stephan Sigg | Physical layer device interaction | 22

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

Virtually all present cryptosystems can theoretically be broken

by an exhaustive key-search

Stephan Sigg | Physical layer device interaction | 23

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

Virtually all present cryptosystems can theoretically be broken

by an exhaustive key-search Probably, they might even be broken due to novel algorithms

Stephan Sigg | Physical layer device interaction | 23

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

Virtually all present cryptosystems can theoretically be broken

by an exhaustive key-search Probably, they might even be broken due to novel algorithms Or by progress in Computer engineering

Stephan Sigg | Physical layer device interaction | 23

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

The One-time pad: An unconditionally secure cryptosystem

(secure against adversary with unbounded computing power)

Message M = [m1, m2, . . . , mN] Key K = [k1, k2, . . . , kN] (uniformly distributed N-bit string) Cipher-text C = [c1, c2, . . . , cN] = [m1 ⊕ k1, . . . , mN ⊕ kN] The one-time pad is perfectly secret

Stephan Sigg | Physical layer device interaction | 24

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

The price we have to pay for perfect secrecy: communicating parties must share secret key that is at least as long as the message and which can only be used once

Stephan Sigg | Physical layer device interaction | 25

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

The price we have to pay for perfect secrecy: communicating parties must share secret key that is at least as long as the message and which can only be used once

Scheme is quite impractical, but...

Perfect secrecy can not be obtained less expensive [Shannon] One-time pad optimal with respect to key length Every perfectly secret cipher is necessarily as impractical as the

• ne-time pad

Stephan Sigg | Physical layer device interaction | 25

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

By using noise we can do better ! The assumption that the adversary has perfect access to the cipher-text is unrealistic in general Signal transmission over a physical channel is subject to noise Utilise noise to achieve a perfectly secure communication at less cost

Stephan Sigg | Physical layer device interaction | 26

Introduction Recognition Calculation Security Conclusion

Fuzzy cryptography

Utilise noise to improve security

Stephan Sigg | Physical layer device interaction | 27

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Shake well before usea

• aR. Mayrhofer et al., Shake well before use: Authentication based on Accelerometer data,

Pervasive 2007

Accelerometer data Secure key by iterative exchange of hashed key-sub-sequences

RF-based ad-hoc secure device pairinga

• aS. Mathur et al., ProxiMate: proximity-based secure pairing using

ambient wireless signals, MobiSys 2011

Utilise error correcting codes Public RF-source

Stephan Sigg | Physical layer device interaction | 28

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

possible messages X possible codewords C

x c′

Decoding Encoding

c

C

Generation of Audio fingerprints3 Utilise Fuzzy cryptography to obtain identical keys at devices4

• 3J. Haitsma and T. Kalker, A highly robust audio fingerprinting system, ISMIR 2002
• 4P. Tuyls, B. Skoric, T. Kevenaar, Security with noisy Data, Springer, 2007

Stephan Sigg | Physical layer device interaction | 29

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Audio-based ad-hoc secure pairinga

• aS. Sigg et al., Secure Communication based on Ambient

Audio, Accepted for IEEE Transactions on Mobile Computing

Audio as common context source Fuzzy cryptography

Clap Music Snap Speak Whistle 0.5 0.55 0.6 0.65 0.7 0.75 0.8

Hamming distance in created fingerprints (loud audio source in 1.5m and 3m)

Audio sequence class

Median percentage of identical bits in fingerprints Fingerprints created for matching audio samples Fingerprints created for non−matching audio samples 2 4 6 8 10 12 14 16 18 20 0.91 0.93 0.95 0.97 0.99 1.01 Test run Percentage of passed tests

Percentage of tests in one test run that passed at >5% for Kuiper KS p−values

1.01947 (confidence value at α = 0.03) 0.92053 (confidence value at α = 0.03) Only music Only whistle Only snap Only speak Only clap

Stephan Sigg | Physical layer device interaction | 30

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Hardware-originated synchronisation offset We experiences significant differences in audio samples from devices with differing hardware(Nexus One; Nexus N) How can we correct these without disclosing information on the channel?

Stephan Sigg | Physical layer device interaction | 31

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

How to synchronise audio without disclosing information?

No data shall be transmitted among devices Hardware-originated synchronisation offset Approximate pattern matching with arbitrary common sequencea

• aT. F. Smith and M. S. Waterman.

Identification of common molecular subsequences. Journal of molecular biology, 147(1):195ˆ a197, Mar. 1981 Stephan Sigg | Physical layer device interaction | 32

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Hardware-originated synchronisation offset Synchronisation in the order of 3ms possible No additional data transmitted among devices

Stephan Sigg | Physical layer device interaction | 33

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Potential and challenges

IoT promotes information-dissemination unattended ’over-the-air’

Such an environment requires unattended, spontaneous ad-hoc security schemes

Stephan Sigg | Physical layer device interaction | 34

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Potential and challenges

IoT promotes information-dissemination unattended ’over-the-air’

Such an environment requires unattended, spontaneous ad-hoc security schemes Sustainable: Reduces cost, processing time and prevails attention No need for additional interfaces No human intervention that slows down a process Human attention not distracted by security-process

Stephan Sigg | Physical layer device interaction | 34

Introduction Recognition Calculation Security Conclusion

Security from environmental stimuli

Potential and challenges

IoT promotes information-dissemination unattended ’over-the-air’

Such an environment requires unattended, spontaneous ad-hoc security schemes Sustainable: Reduces cost, processing time and prevails attention No need for additional interfaces No human intervention that slows down a process Human attention not distracted by security-process Potential: Other context sources and improving security RF-channel features, Light, proximity Generalise RF-features for more than two devices Improve authentication for RF-security schemes

Stephan Sigg | Physical layer device interaction | 34

Introduction Recognition Calculation Security Conclusion

Conclusion

Do you have any questions?

Stephan Sigg sigg@nii.ac.jp

Stephan Sigg | Physical layer device interaction | 35