Embed Size (px)
Citation preview
A Survey of Wearable Devices Pairing Based onBiometric Signals
Jafar Pourbemany, Ye ZhuDepartment of Electrical Engineering and Computer Science
Cleveland State UniversityCleveland, USA
[email protected], [email protected]
Riccardo BettatiDepartment of Computer Science and Engineering
Texas A&M UniversityTexas, USA
Abstract—With the growth of wearable devices, which areusually constrained in computational power and user interface,this pairing has to be autonomous. Considering devices thatdo not have prior information about each other, a securecommunication should be established by generating a sharedsecret key derived from a common context between the devices.Context-based pairing solutions increase the usability of wearabledevice pairing by eliminating any human involvement in thepairing process. This is possible by utilizing onboard sensors(with the same sensing modalities) to capture a common physicalcontext (e.g., body motion, gait, heartbeat, respiration, and EMGsignal). A wide range of approaches has been proposed to addressautonomous pairing in wearable devices. This paper surveyscontext-based pairing in wearable devices by focusing on thesignals and sensors exploited. We review the steps needed forgenerating a common key and provide a survey of existingtechniques utilized in each step.
Index Terms—Pairing, authentication, wearable device, net-work security, body area network, spontaneous pairing, security,biometric, internet of things, sensor.
I. INTRODUCTION
Smart wearable devices can collect a variety of informationabout human activities and behaviors which makes them pop-ular in clinical medicine and health care, health management,workplace, education, and scientific research. The wearablemarket is diversified with hundreds of products, includingsmartwatches, smart wristbands, smart glasses, smart jewelry,smart straps, smart clothes, smart belts, smart shoes, smartgloves, skin patches, and even implanted medical devices(IMD) [1]–[8]. As Fig. 1 illustrates, there are many diversetypes of wearable devices for different body parts that col-lect various physiological data and possibly make intelligentdecisions based on the collected physiological data. Wearabledevices are exploited in a wide range of applications [6]. Forexample, wearable devices can be used as health monitoringsystems and health treatment systems. They can monitorvital signs like heart rate [9]–[11], respiratory rate [12]–[15],body temperature [16], [17]. Others collect parameters likeblood pressure [18], blood oxygen [19], blood glucose [20],to detect disorders. Some wearable devices also can helpdisabled patients to recover certain physical functions [21]–[23]. Wearable devices are applied beyond healthcare. Forexample, wearable devices can be used to recognize dailyphysical activities [24]–[30] or the activities that are related to
specific sports [31]–[47]. Other applications areas for wearabledevices include payment management [48], [49], unlockingvehicles [50], keeping sensitive information (like passwords)[50], controlling paired devices [51], subject tracking [52],[53], fall detection [54]–[58], drowsiness detection [59]–[61],environment monitoring [62], [63], virtual and augmentedreality [64], [65], and spying [49].
Based on their application, the wearable devices collect,analyze, and store the data. In some cases, the wearable device(e.g., pacemaker) may be equipped with actuators which itmay use to control important physiological functions of thewearer. Often, they need to share data or commands with thebase station or other wearable devices. Data/command sharedbetween wearable devices are usually sensitive like stored cre-dentials of the user or commands to adjust an IMD. Therefore,the communication should be protected by encryption and thedevices need to have a common key for encryption to ensurethat the process has not been compromised by an attacker.
Many wearable devices use Bluetooth or Wi-Fi to connect toeither other smartphones or other wearables, and the Internet.To be able to send and receive data the new wearable devicemust establish a connection with the other device; this processis called pairing (also known as binding, coupling, bonding,or association [3]). For example, the user initially needs topair a new smartwatch with their smartphone over Bluetoothor NFC channel before use. Some wearable medical devicessuch as IMDs need to be paired to their external controllersto receive updates.
Traditional pairing techniques often need user actions canhave various forms. A common approach is used in Bluetoothpairing: The user selects a target device from a list of availabledevices and then possibly uses a PIN for additional authentica-tion [66]. However, the approaches that require an initial stageof network setup are not scalable as the number of wearablesincreases [67]. PIN-based approaches need interaction with thedisplay, which may either be inconvenient or even impossiblein many wearables [67], [68]. Furthermore, PIN-based pairingis vulnerable to observation attacks such as shoulder surfing[67]. Public key cryptography (PKC), on the other hand,cannot be used to create a secure key on wearable devicesbecause it requires a public key infrastructure (PKI) [69].In addition, it requires expensive computing methods that
arX
iv:2
107.
1168
5v1
[cs
.CR
] 2
4 Ju
l 202
1
are not suitable for resource-limited IoT devices. To mitigatethese limitations, various techniques have been proposed thattake advantage of common features in different wearabledevices. Pairing based on biometric signals is a natural choicebecause wearable devices are attached to the same body. Bothbehavioral biometrics (e.g., step counts) and physiologicalbiometrics (e.g., heart rate) are used for wearable devicepairing and authentication.
In this article, we survey context-based techniques that canbe used to achieve automatic wearable pairing based on them.In Section II, we review a selection of survey papers onbiometric-based approaches. Section III discusses the signalsused for pairing smart wearables. In Section IV, differentsteps of biometric-based pairing mechanism are surveyed.In Section V, we investgate the most common adversarymodels. Limitations and challenges are presented in SectionVI. Finally, Section VII concludes the paper.
II. RELATED STUDIES ON BIOMETRIC-BASED APPROACHES
Biometric signals are widely used for device authentica-tion. Variety of context-based approaches are used in deviceauthentication to verify user identity [70] on a unique device.
There are several surveys that have investigated differentaspects of the security and authentication in the wireless bodyarea network (WBAN) [71], [72]. In Table I, we summarizethe topics raised in surveys over the last ten years. Securityessentials and existing attacks against WBAN have been ad-dressed in Javadi, and Razzaque [73]. They also discussed thesignificant constraints and challenges of security mechanisms.A detailed review of authentication schemes is conducted byMasdari and Ahmadzadeh [74]. The authors provided a taxon-omy for authentication modalities that classifies it into threecategories: biometric-based, channel-based, and cryptography-based. Mainanwal et al. [75] summarized the pros and consof several security and privacy techniques and discussed thethreats and challenges. Sawaneh et al. [76] surveyed theviews on security and privacy essentials for WBAN. Keystrokedynamics are addressed in [77], [78]. Abuhamad et al. [79] in-vestigated the authentication protocols and OS-related securityof smartphones’ users using behavioral biometrics.
The survey by Janabi et al. [80] reviewed the major securityand privacy problems in WBAN for healthcare applications,along with their state-of-the-art security solutions. The paperprovides significant highlights on open issues and futureresearch directions in WBANs. Aman and Shah [81] exploredthe security problems of ubiquitous healthcare (U-Healthcare)related work. The survey by Narwal and Mohapatra [82]focuses on the analysis of authentication schemes in termsof main outcomes, strengths, and limitations. In addition tothis, the authors discuss the architecture of WBAN, securityessentials, and security attacks are discussed in detail. Ataxonomy is proposed by Usman et al. [83] that classifysentities involved in healthcare systems. Security challenges atall WBAN tiers have been studied. The authors also identifiedopen issues and highlighted future research directions. Inanother work, the cryptographic solutions have been reviewed
by Malik et al. [84]. They provided a general survey on majorsecurity essentials and conceivable assaults at different layers.
Considering security view of the complete WBAN system,Morales et al. [85] focused on various protocols in WBANarchitecture and provided a detailed review of security re-quirements such as confidentiality, integrity, privacy, authenti-cation, and authorization. Komapara and Holbl [86] reviewedsecurity and key agreement of Intra-BAN communication.They categorised the existing key agreement schemes intotraditional, physiological value-based, secret key-based, andhybrid key-based schemes. Furthermore, the authors providedthe description of each class and analysed the security strengthof BAN towards attacks. Nidhya and Karthik [87] emphasizedthe security attacks and security models at data collection,transmission, and data storage level. Also, they assessed theprivacy requirements and reliability of healthcare systems.Joshi and Mohapatra [88] investigated design, functionalities,and workflow of the existing authentication schemes. Theydescribed the structure of authentication schemes and provideda detailed view on communication standard and design issuesin WBAN. The paper also suggests methods to safeguard thekey during key management.
Surveying security and Privacy issues in WBAN, Chaudharyet al. [89] classified authentication schemes into four cate-gories: Physiological Value-based, Channel-based, Proximity-based, and Cryptographic-based. Furthermore, they summa-rized various schemes from different categories in a tabularform to highlight the features of each scheme effectively.The survey by Hussain et al. [90] conducted to providegreater insight into authentication schemes. All in all, thesurvey presented a detailed discussion on security features,security attacks, strengths, limitations, and performance of theauthentication schemes. Roy et al. [91] reviewed the majorsecurity and privacy issues in WSNs and WBANs. Theyconducted a comparative analysis of both the networks basedon their features, architecture, applications and threats. Inanother work, Narwal et al [92] explored the authenticationschemes in different categories.
All the aforementioned studies surveyed a variety of meth-ods used sensory signals to assist IoT device authentication.However, sensors data, especially biometric signals, can beused for pairing the wearable devices as well. Indeed, bio-metric signals are the common point between wearable deviceauthentication and pairing, whereas the goals and applicationsare different. This survey focuses on sensor-based pairing inwearable devices to provide more details about the challengesand limitations of sensors in such devices.
III. SIGNALS USED FOR PAIRING
Various types of sensors collect a variety of informationfrom the human body and the environment. As Fig. 2 shows,several signals are used for context-base pairing, includingmotion, gait, ECG, PPG, EMG, and SCG. Indeed, thesesensors can provide auxiliary out-of-band (OOB) channels [96]as a feasible option to facilitate device pairing. We categorize
Fig. 1. Different types of wearable devices.
and describe the signals used for wearable device pairing inrecent papers in this section.
A. Motion and position
Magneto-Inertial Measurement Unit sensors (MIMU), in-cluding accelerometer, gyroscope, and magnetometer, are themost common sensors in today’s wearable devices. Thesesensors detect the user’s motion and heading of the device withrespect to the Earth’s magnetic north pole. The accelerometermeasures acceleration in three orthogonal spatial dimensions,x, y, and z, where each axis denotes either the vertical,forward-to-backward, or left-to-right dimensions in meter persecond squared [97]. The gyroscope measures the angularrotation about each of these axes in radians per second [98].The magnetometer measures the strength of the local magneticfield along three orthogonal axes [99]. User’s body movementcan be modeled by using the information provided by thesesensors. Hence, a variety of methods have been proposed touse such sensory data for authentication and pairing purposes.
For instance, in [100], [101], the authors proposed to shakethe two mobile devices held in one hand; in this way, theaccelerometer reading could be used to generate matched keys
on either frequency domain [100] or time-domain [101]. Theauthors demonstrate that the simultaneous shaking motion oftwo devices generates unique accelerometer readings that anadversary cannot easily mimic at a close distance. Shen et al.[102] proposed a method in which using a similar motion pat-tern of handshaking, two devices on the different bodies can bepaired. Similarly, in [103], a handshake-based pairing schemebetween wrist-worn smart devices are developed based on theobservation that, by shaking hands, both wrist-worn smartdevices conduct similar movement patterns. Hash functionsand heuristic search trees were leveraged in [104] to proposea key exchange protocol based on accelerometer data whilethe user shakes devices together. In another work, Yuzuguzelet al. [105] propose to derive a shared key based on finding asmall set of effective features from shaking of two devices thatare held together in one hand. Gorza et al. [106] investigatedthe pairing of mobile devices based on shared accelerometerdata under various transportation environments. Using multiplesensors (accelerometer and microphone), Cao et al. [107]presented a device-to-Device (D2D) communication that theuser needs to hold two devices in one hand and randomlyshake them for a few seconds. Jin et al. [99] proposed a
TABLE ISURVEY PAPERS ON THE WBAN AUTHENTICATION - CONTENT COMPARISON
Scheme Year Methodsclassifications
Knowledge-basedmethods Motion Gait ECG PPG Advantages and
disadvantagesAttacksenarios
Futureworks
[78] 2011 [73] 2013 [77] 2013 [75] 2015 [74] 2016 [76] 2016 [93] 2016 [94] 2016 [95] 2016 [81] 2017 [80] 2017 [82] 2018 [83] 2018 [84] 2018 [85] 2018 [86] 2018 [87] 2019 [88] 2019 [89] 2019 [90] 2019 [91] 2020 [92] 2021 [79] 2021
scheme to pair smartphones in close distance by exploitingcorrelated magnetometer readings.
Table II shows a summary of features of several papers inthe field of motion and position based device pairing.
B. gait
Gait recognition is the process of identifying an individualvia how he or she walks based on wearable sensors data,especially motion sensors (e.g., accelerometer and gyroscope)[110]. Due to the different properties of an individual’smuscular-skeletal structure, gait patterns are fairly uniqueamong individuals [111]. Hence, it can be determined if twodevices are carried by the same person [112].
Various techniques exploit different features of gait togenerate a common key for pairing wearable devices. Sunet al. [113] proposed a method to generate a symmetric keybased on the timing information of gait. The authors usedthe IPI of consecutive gait as a common feature betweenthe two devices. Schurmann et al. [114] presented a securespontaneous authentication scheme that exploits correlation inacceleration sequences from devices worn or carried togetherby the same person to extract always-fresh secure secrets. Intheir method, BANDANA, they utilized instantaneous varia-tions in gait sequences with respect to the mean. Walkie-Talkie[115] is another shared secret key generation scheme thatallows two legitimate devices to establish a common crypto-graphic key by exploiting users’ walking characteristics (gait).The authors exploit independent component analysis (ICA)for blind source separation (BSS) to separate accelerometer
signals from different body movements such as arm swing andwalk. In Gait-Key [116] Xu et al. extended their method inWalkie-Talkie to examine the effect of multi-level quantizationon the pairing success rate. In [117] the same authors alsoproposed using spatial alignment instead of using BSS. Ausability analysis of four gait-based device pairing schemes[104], [113]–[115] are presented in [118]. A summary offeatures of several gait-based pairing papers is shown in TableIII.
C. ECG and PPG
The heart-beat is a promising option for wireless bodyarea networks (WBANs) authentication and key generatingschemes because its properties are unique, and their featuresdiffer from person to person [119]. Heart-beat signals can beeasily collected, and they are hard to copy by other peoplein comparison to simple pin codes. It is more secure thantraditional methods because it requires a user to be availableat the time of authentication and pairing process [120], [121].Heart-beat signals can usually be collected by ECG and PPGsensors. ECG sensors collect the electrical activity of heartmuscles through electrodes attached to the body. PPG sensorswhich can be attached to different parts of the body likethe ear and finger, detect the blood level transforms in themicrovascular cot of tissue [122]. It illuminates the body andmeasures transforms in light absorption as blood circulatesin the body. The heart-beat signal can also be measured byseismocardiogram (SCG), which is the chest movement inresponse to the heart-beat. Accelerometers and piezo vibration
Fig. 2. Signals used for pairing.
sensors in the wearable devices can measure SCG as well[123]–[125].
Various features extracted from heart-beat signals can beused for authentication and key generating purpose. The mostimportant feature used in WBANs is heart rate variability(HRV) or R-R interval or inter-beat interval (IBI) or Inter-pulseInterval (IPI) [126]–[128] indicates the time interval betweenconsecutive heart-beats [129]. Indeed, the fluctuations of heartrate around an average rate are shown by HRV [126]. As hasbeen proven by several studies [130]–[132], HRV is highlyrandom and can be used as a random source to generate keys.Since HRV is a unique characteristic for each person, it canbe used as an authentication method to pair devices on thesame body.
Rostami et al. [131] proposed an HRV-based pairing methodto authenticate external medical device controllers and pro-grammers to IMDs. The authors introduce a touch-to-accesspolicy using a time-varying physiological value (PV) by ECGreadings. They utilized statistical characterization of ECG forpairing wearable devices. Another pairing system called H2Bis presented by Lin et al. [123], which utilizes piezo sensorsto detect heart-beat signals and generate a secret key.
D. EMG
The EMG or electromyogram signals are the electricalsignal generated by contractions of human muscles. Accordingto medical research [133], [134], the EMG signal is a quasi-random process, i.e., the average value of EMG is correlatedto the generated force of the muscle, but it has a random am-plitude variation under a given force. In other words, there arestochastic variations of EMG amplitude for a unique gestureand force. Therefore, the EMG signals can be used as a securesource to generate secret keys in physically close contact forsome wearable devices like Myo armband [135], Athos gear[136], and Leo smart band [137]. Since detecting this kindof signal needs physical contact in proximity, it is extremelydifficult for an adversary to perform an eavesdropping attack.EMG-KEY is an EMG-based method proposed by Yang etal. [138] which leverages EMG variation signal to generate asecret key for pairing two wearable devices.
A summary of different signals and sensors used for wear-able device pairing is shown in Table IV.
IV. BIOMETRIC-BASED PAIRING MECHANISM
As Fig. 3 shows, sequence of signal processing is neededto separate sources (signal and noise), detect and extract
TABLE IIMOTION-BASED PAIRING PAPERS AND THEIR FEATURES
Sche
me
Year
Sens
oran
dTo
ols
Key
Len
gth
(bit)
Dur
atio
n(s
ec.)
Succ
ess
Rat
e
Equ
alE
rror
Rat
e(E
ER
)
FAR
FRR
Bit
Agr
eem
ent
Com
puta
tion
Tim
e
Ene
rgy
Con
sum
ptio
n
Freq
uenc
y
Subj
ects
Dev
ice
[102] 2018 Acc 128 1 >99% 1.60% X X X X Wrist worn smart wear-ables
[108] 2020 Acc 128 2.33 X X 25 LaunchPad CC1350[103] 2019 Acc 128 close
to100%
1.5% X X X 100 16 (8m 8f) iPhone6 and iPhone8
[106] 2020 Acc 448 90 X 5 LG Optimus L7 P700,Samsung J5
[107] 2019 Acc,Mic,Speaker
<5% <2% >90% X X 100 10 (6m 4f) HUAWEI MATE8
[105] 2015 Acc 40 5 76% 4% X X 100 10 (5m 5f)[109] 2020 Acc, Vi-
bration12.5 1 85.90% 5.00% 500 12 (5m 7f) Arduino, accelerometer
MPU-6050[104] 2012 Acc 64 75% 80-
90WiiMotes
[100] 2007 Acc 128 3 X X 600 51 (32m 19f) ADXL202JE accelerome-ters
[101] 2007 Acc 140 80% 200 10[99] 2016 Magnet-
ometer128 4.5 >=
90%X 50 Various kinds of smart-
phones
TABLE IIIGAIT-BASED PAIRING PAPERS AND THEIR FEATURES
Sche
me
Year
Sens
oran
dto
ols
Key
leng
th(b
it)
Dur
atio
n(s
ec.)
Succ
ess
rate
Bit
agre
emen
t
Com
puta
tion
Tim
e
Ene
rgy
Con
sum
ptio
n
Freq
uenc
y
Subj
ects
Dev
ice
[113] 2017 Acc 79% X 100 5 iPhone[114] 2018 Acc 16 12 >=75% 50 482 and 15[115] 2016 Acc 128 5 X X X 100 20 Motorola E2[116] 2017 Acc 128 4.6 98.30% X X X 100 20 (14m 6f) Motorola E3[117] 2019 Acc 128 5 X X X 100 21 (14m 6f) Motorola E4
features/events, quantize features/segments, correct errors, am-plify bit string and create a shared key in the different devices.We explain these sequences in the following paragraphs.
A. Data acquisition and preprocessing
1) Sampling: Sampling is the basic stage in which raw datais collected through various sensors, from the accelerometerto the ECG sensor. The speed and quality of data acquisitioncan directly affect key generation. Ambient noise and systemnoise cause the bit current of both devices to mismatch andlead to unsuccessful pairing. On the other hand, the amountof sampling affects the amount of entropy extracted from thesensors’ values. This is because the sampling rate determinesthe accuracy of the measurement: the higher the sampling rate,
the more accurate the measured signal, the more noise [139].Therefore, as the sampling rate increases, the measurementswill have more entropy.
2) Synchronization: Synchronization is required to ensureall legitimate devices capture samples at the same time. Sincedevices are unsynchronized by default, they must agree on acommon starting point. Time asynchrony would result in ahigh bit string mismatch rate between devices after bit quan-tification, which would decrease the probability of successfulkey distribution. Since both smartphones are unsynchronizedinitially, they need a user interaction for synchronizing thestarting points of recording the shaking process. This taskcan be realized through direct user input, such as pressinga button. Alternatively, synchronization can be done by using
TABLE IVDIFFERENT TYPES OF SIGNALS AND SENSORS USED FOR PAIRING
Scheme Name Year Signal Sensorandtools
Description Technique Attack senarios
[108] SHAKE 2020 Motion Acc Holding devices together and shaking them.SHAKE leverages the shared accelerationpatterns to generate the same key in bothdevices.
Shaking together Eavesdroping,mimic attack
[106] Multi-ModalTransport
2020 Motion Acc Pairing of mobile devices based on ac-celerometer data under various transporta-tion environments
Detecting the same varia-tion in different devices
Full control of thecommunicationchannel
[109] VibeRing 2020 Motion Acc, vi-bration
Use of vibration, generated by a customRing, as an out-of-band communicationchannel to unobtrusively share a secret witha Thing
Detecting vibration of ring Impersonating,eavesdropping
[103] Shake toCommu-nicate
2019 Motion Acc A secure handshake acceleration-based pair-ing mechanism for wrist worn devices
Handshaking Passive attack, ac-tive attacks
[107] Sec-D2D 2019 Motion Acc,Mic,speaker
Device-to-Device (D2D) communication byusing multiple sensors, acceleration and mi-crophone by holding two devices in onehand and randomly shaking over a periodof time
Shaking together Impersonation
[117] 2019 Gait Acc Gait-based shared key generation systemthat assists two devices to generate a com-mon secure key by exploiting the user’sunique walking pattern
Using rotation matrix Eavesdropping, ad-versary
[123] H2B 2019 SCG PiezoVibra-tionSensors
Using piezo sensors to detect heartbeat sig-nal and generate a secret key
Detecting piezo-based IPI Passive eavesdrop-ping attacks, Activepresentation attacks,Active video attacks
[102] Shake-n-Shack
2018 Motion Acc Enabling secure data exchange betweensmart wearables via handshakes
Handshaking Mimicking attacks
[114] BANDANA 2018 Gait Acc A secure spontaneous authenticationscheme that exploits correlation inacceleration sequences from devicesworn or carried together by the sameperson to extract always-fresh securesecrets.
Utilize instantaneous vari-ations in gait sequenceswith respect to the mean
MITN, Mimic Gait,Video Recording,Attach MaliciousDevice
[113] 2017 Gait Acc Symmetric key generation scheme based onthe timing information of gait
Inter-pulse Intervals (IPI)of consecutive Gait
[116] Gait-Key 2017 Gait Acc A shared secret key generation scheme thatallows two legitimate devices to establisha common cryptographic key by exploitingusers’ walking characteristics (gait)
Using independent com-ponent analysis (ICA) forBlind Source Separation(BSS)
Impersonationattack, passiveeavesdroppingadversary, activespoofing attack
[99] MagPairing 2016 Motion Magnet-ometer
Pairing smartphones in close proximity byexploiting correlated magnetometer read-ings
Detecting correlated mag-netometer readings
Passive attacksMITM attacks,replay attacksreflection attacks
[115] Walkie-Talkie
2016 Gait Acc A shared secret key generation scheme thatallows two legitimate devices to establisha common cryptographic key by exploitingusers’ walking characteristics (gait)
Using independent com-ponent analysis (ICA) forBlind Source Separation(BSS)
Impersonationattack, passiveeavesdroppingadversary, activespoofing attack
[138] SecretfromMuscle
2016 EMG EMG A system that can securely pair wearabledevices by leveraging the electrical activitycaused by human muscle contraction, Elec-tromyogram (EMG), to generate a secretkey
Detecting muscle contrac-tion
Impersonation
[105] ShakeMe 2015 Motion Acc Finding a small set of effective featuresfrom shaking of two devices which are heldtogether in one hand
Shaking together
[131] H2H 2013 ECG ECG A system to authenticate external medicaldevice controllers and programmers to Im-plantable Medical Devices (IMDs)
Detecting ECG-based IPI Active adversaries,MITM
[104] SAPHE 2012 Motion Acc Shake devices together, key exchange pro-tocols based on accelerometer data that useonly simple hash functions combined withheuristic search trees
Shaking together MITM
[100] ShakeWellBeforeUse
2007 Motion Acc A method for device-to-device authentica-tion that is based on shared movement pat-terns which a user can simply generate byshaking devices together.
Shaking together MITM, online at-tack , offline attack
[101] 2007 Motion Acc Establish a secure connection between twodevices by shaking them together
Shaking together
Fig. 3. Context based pairing steps.
a coordinator server [102], [140]. Hence, upon receiving thecoordinator’s synchronization signal, all the sensor nodes inthe same network will start recording data [113]. However,we cannot ensure this kind of synchronization is alwaysavailable for users. The wireless communication technique alsocan be exploited for synchronization purposes, e.g., a time-slotted channel hopping-based (TSCH) communication leadsto temporal alignment between devices [141]. Synchronizationcan be at a sample level, i.e., within less than half thesample width, or at an event level, i.e., based on the onsetof detected (explicit or implicit) events with the respectivedevice. Depending on the sensors used for pairing, differentevents can be considered as anchor points. For instance, theevent can be a heel-strike [116], bumping the devices orshaking them together in one hand [105]. Although bettersynchronization reduces the bit mismatches at the next stage,the pairing protocol should support devices without a screenand keyboard and not impose strict synchronization on pairingdevices when collecting ambient context information [138].
3) Filtering and Normalization: The sensor data must passa filtering step to eliminate the relative effect of ambient andartificial noise and extract the desired signal (motion, gait,PPG, ECG, EMG, etc.) from the raw data. Depending on thesignal type, the filtering stage involves using a high-pass filter,a mid-pass filter, or a low-pass filter. For best results, the cut-off frequency of these filters should be in the range of theminimum and maximum frequency of the desired signal; forexample, 3 Hz can be a suitable cut-off frequency for a low-pass filter for gait signal because the normal step frequency liesbetween 1.6-2.8 Hz [117]. In fact, by filtering the raw signal,the unwanted frequency components are removed. Since themagnitude ranges of sensors are quite different, after filtering,the signal is usually normalized to have a mean of zero and avariance of one.
B. Feature extraction
Similar raw signals lead to similar feature signals. Thefeature extraction process maintains the user’s desired char-
acteristics (for example, heartbeats IPI, respiration rate, gait,movement) and eliminate irrelevant noises [142]. Featureextraction is one of the most important steps in a pairingthat significantly affects the bit generation process and itsperformance. In particular, selecting the appropriate featuresplays a key role in pairing because authentication is done bycomparing the bits generated based on the extracted features.Features can be selected from both time and frequency do-mains [143]. Time domain features can be a wide variety offeatures including statistical characteristics [144]–[146] suchas mean; standard deviation; variance; median; root meansquare; maximum; minimum; Skewness; kurtosis; crest fac-tor; number of peaks; Peak-to-peak amplitude; zero crossingrate [147]; signal magnitude area [148]; interquartile range[149], [150]. In the frequency domain we can select featureslike energy, entropy [151], maximum frequency index, meanfrequency, fast Fourier transform coefficients [152], to namebut a few.
C. Quantization
Quantization represents a source’s output with a large (pos-sibly infinite) alphabet with a small alphabet. It is a many-to-one mapping and, therefore, irreversible. Quantization can bedone on a sample-by-sample basis, or it can be performed ona group of samples (by dividing the signal into segments).Quantizer encodes each sample/segment by specifying therange value at multiple levels. In this process, the raw signalsor features will be quantized into bit vectors. The quantizerhas a significant role in the security of the generated key.By making a biased quantization, a brute force attack wouldbecome feasible. On the other hand, the quantizer can changethe length of the final key. The quantizer maps each sample toa bit string whose length depends on the quantization levels. Aquantizer that has 2m quantizing levels can map each sampleto m bits [139].
The bit representation can also affect security. The quantizercan map samples to either binary code or Gray code [153]. AGray code encodes numbers so that adjacent numbers have
a single-digit differing by one. Therefore, in some cases,the Gray code can perform better in detecting noise in twoconsecutive samples [131]. In the recent pairing methods,various type of quantization approaches has been exploited,including pairwise nearest neighbor (PNN) quantization [101],standard decimal-to-binary quantizer [99], [105], decimal-to-Gray-code quantizer [123], [131], uniform quantizer [113],sigma-delta quantizer [106], and exploiting multiple thresholds[100], [102]–[104], [107]–[109], [114], [115], [117].
D. Error Correction
Due to the measured noise, there are usually mismatches inthe quantized bits between the bit vectors produced by thetwo devices. Therefore, in the reconciliation stage, devicesexchange a certain amount of information to correct all mis-matches and generate a bit-by-bit matching key. We describesome of the most important error correction approaches usedin wearable device pairing in the following.
1) Index checking: This technique exchanges the indexof the valid bit positions to reach a mutual agreement onwhich bits will be used in the final keys. For example,suppose the key generated by Alice’s devices is [110xx11x00],while for Bob is [1100x11xx0], where x means the posi-tion where no valid bit is presented. Then both Alice andBob inform each other the positions of the valid bits, i.e.,Alice sends PAlice = 1, 2, 3, 6, 7, 9, 10, and Bob sendsPBob = 1, 2, 3, 4, 6, 7, 10. Upon receiving the positions, theycompare the received vector with the local one and agreethat only the bits that are valid according to both vectorsshould be used. In this example, the agreed positions should be1, 2, 3, 6, 7, 10 so that the final symmetric keys are [110110].This error correction technique is utilized in [102], [115],[117].
2) Error Correction Code (ECC): Error correction code(ECC) is commonly used to control data errors throughunreliable or noisy communication channels [154]. The centralidea is that the sender encodes the message with additionalinformation in an ECC form. The redundancy allows thereceiver to detect a limited number of errors that may occuranywhere in the message, and often to correct these errorswithout retransmission [155]. Bose–Chaudhuri–Hocquenghem(BCH) [156], Reed-Solomon (RS) [157], Hamming [158], andbinary Golay Codes [159] are some of the most common ECCused in the recent pairing methods [107], [109], [113], [114],[116], [138]. BCH codes form a class of cyclic error-correctingcodes constructed using polynomials over a finite field (alsocalled Galois field). One of BCH codes’ key features is thatthere is precise control over the number of symbol errors thatcan be corrected by the code when code design. Specifically,it is possible to design binary BCH codes can be designed thatcan correct multiple bit errors [160]. Reed-Solomon codes arethe subset of BCH codes among the most powerful knownclasses of linear, cyclic block codes. Reed Solomon describes asystematic way of building codes that could detect and correctmultiple random symbol errors. By adding t check symbolsto the data, RS code can detect any combination of up to t
erroneous symbols or correct up to t/2 symbols. In addition,RS codes are suitable as multiple-burst bit-error correctingcodes because a sequence of b+ 1 consecutive bit errors canaffect up to two symbols of size b [161]. Hamming code is ablock code that can detect up to two simultaneous bit errorsand correcting single-bit errors. Binary Golay code is anotherlinear error-correcting code in which a codeword is formed bytaking 12 information bits and appending 11 check bits.
3) Fuzzy cryptography: The fuzzy cryptographic schemeenables the compatibility of a certain amount of tolerablenoise between the keys extracted from different devices bychanging the error correction parameters and the length ofthe samples used. ”Fuzzy,” in this context, refers to thefact that the fixed values required for cryptography will beextracted from values close to but not identical to the originalkey, without compromising the security required [162], [163].Jiang et al. [103] used this technique for error-correcting intheir pairing algorithm. In [113], Sun et al. exploited fuzzycryptography and BCH to provide the superior performanceof false acceptance rate (FAR).
4) Compressive Sensing: Compressed sensing theory hasshown that sparse signals can be reconstructed exactly fromremarkably few measurements [164]. Hence, it can be ex-ploited as an error correction method [165], [166]. Compressedsensing is a technique where a signal x is multiplied byan M × N(M < N) sampling matrix to be sampled andcompressed in a single operation. The signal x is recoveredby finding the l1 norm of the sparse version of x, representedby the received samples y. In other words, out of the infinitesignals which could have been used to create the received y,the sparsest one is recovered as the original signal. As longas x is ”sparse enough,” it can be recovered exactly using thistechnique. Lin et al. used this method in their error correctionmethod [123].
E. Key Amplification
In the reconciliation phase, the devices may send someinformation to each other through a public wireless chan-nel. Besides, some slightly different samples/segments mayhave the same bit string due to error correction. Thus, anadversary can infer some private information about the secretsequence. The privacy amplification process solves this issue.Key amplification helps to increase the final key’s randomnessto eliminate the information leakage and increase entropy.Typically, two methods are used to combine keys generatedfrom different segments and eliminate the correlation betweenthem, bitwise XOR function [115]–[117] and hash functionMD5 and SHA-256 [100], [101], [107], [108], [114], [131].
F. Performance metrics
The ultimate objective of pairing is to generate the samekey on different devices independently. On the other hand,due to wearable devices’ hardware limitations, all the oper-ations in such devices are expected to use as little memoryand computational power as possible and communicate thesmallest amount of data with the smallest number of messages
to achieve the smallest overall energy consumption. Thereare several evaluation metrics for evaluating pairing systemperformance which we have provided a brief description ofthem in the lines below [79], [86], [167].
• Bit generation rate:The number of bits generated from the sensor readingsper second.
• Key generation rate:The major performance metric for symmetric key gen-eration is the success rate, the probability that two keysgenerated by Alice and Bob can completely agree witheach other (The probability of 100% matching). In otherword, it is the percentage of identical keys generated bytwo devices in one second.
• Bit agreement rate:Bit Agreement Rate denotes the percentage of the match-ing bits of the two cryptographic keys generated by twodevices
• Computational cost:The next important performance indicator is computa-tional cost. It is important for schemes to be as computa-tionally efficient as possible, because sensor nodes do notpose much processing power and because more comput-ing uses up more of the very much limited energy supply.The most common method to analyze computation cost isby measuring the amount of time it takes for the necessaryoperations to finish processing.
• Energy consumption:energy consumption was measured by how much energyis spent on every bit of information produced.
• Communication cost:Measuring of communication cost is very important,because it is the most energy consuming operation ofthem all. The most common ways of determining thecommunication cost are by the size of the sent data
• False positive ratio:False positive ratio (FPR), also known as false accept rate(FAR), is defined as the percentage of pairing attemptsthat incorrectly generate common keys among all theexpected unsuccessful attempts. The metric indicates thelikelihood of the adversary successfully pairing with alegitimate device by the adversary. The FPR can becomputed as:FPR = FP
EN , where FP is the number ofincorrectly generated keys, and EN is the number ofexpected unsuccessful attempts.
• False negative ratio:False negative ratio (FNR), also known as false rejectrate (FRR), is defined as the percentage of incorrectlyunsuccessful pairing attempts among all the expectedsuccessful pairing attempts. It indicates the probabilitythat two legitimate devices attached to the same body cannot pair successfully. FNR is computed as FNR = FN
EP ,where FN is the number of incorrectly missed keys, andEP is the number of expected successful attempts.
• Equal/crossover error rate (EER/CER):
Equal or crossover error rate (EER/CER) is the rate atwhich both acceptance and rejection errors are equal.The value of the EER can be easily obtained from theintersection point between the FAR and the FRR curves.Equal Error Rate (EER) measures the trade-off betweenFAR and FRR and it is the value of FAR or FRR whenthe two false rates are equal.
• Entropy:Another important security metric in the pairing schemesis the entropy estimation. Entropy is the measure ofuncertainty or randomness in the bit string generatedfrom the measured signals. Higher entropy means morerandomness to the generated bit string or in other wordsless dependencies between the bits. Some papers use theNIST test suite [168] to estimate the entropy.
A list of various processes and techniques used in wearabledevices pairing techniques is shown the Table V.
V. ADVERSARY MODEL/ ATTACK SCENARIOS
Wearable device pairing can face various attacks largelydue to the broadcast nature of the wireless communicationbetween wearable devices. We must consider the presence ofa strong attacker during key generation. Eve is fully awareof the system and control of the communication channel,meaning that she may monitor, jam, and modify messagesat will. Various attacks on wearable devices pairing can becategorized into passive and active attacks [92], [169], [170].In this section, we provided a brief description of these attacksand discussed some common countermeasures.
A. Passive attacks
A passive attacker monitors the network for informationbut does not affect the target network. Passive attacks areperformed as a preliminary act for the active attack [171].Data Sniffing/Eavesdropping: Data Sniffing or Snooping attackis an old security issue. Sniffing is an incursion that involves aweak connection between the WBAN node and the server. Theattacker passively accesses the data traffic (important healthdata, routing updates, node ID numbers, etc.) for later analysisby sitting between the unsecured network paths. Detectinga passive attack is exceedingly difficult and impossible inmany cases because it does not involve any changes. However,protective measures can be implemented to stop it, including:Avoid posting sensitive information publicly, using randomkey distribution and strong encryption techniques to scramblemessages, making them unreadable for any unintended recip-ients.
B. Active attacks
An active attack involves using information gathered duringa passive attack to compromise a user or network. Activeattackers can cause devastation to the system as they attemptto intercept the wireless communication to change the infor-mation present on the target or en route to the target. There areseveral types of active attacks. In an impersonation/spoofingattack, an attacker pretends to be another user to access
TABLE VPROCESSES AND TECHNIQUES UTILIZED IN WEARABLE DEVICE PAIRING STUDIES
Scheme Synchronization Filtering Feature Extrac-tion
Quantization BitPresen-tation
Error Correc-tion
Amplifi-cation
Confirmation Encryp-tion
[102] First peak Dominantmotion features,PCA
Thresholds Binary Index exchange AES
[108] Using gateway Low-passfilter 2 Hz,mean filter
Hysteresisthreshold-ing
Binary Error correctioncode
Hashfunction
Challenge-Response(CR)mechanism
[103] First samplewith magnitudeof 0
Low-pass fil-ter 5Hz
Accelerationmagnitude of 0
Thresholds Binary Fuzzy cryptog-raphy
MAC(messageauthentica-tion code)
[106] Exchange timemessage byan unsecurechannel
Low-passfilte andHigh-pass
Sign of eachsample
Sigma-delta Binary EKEandSPEKE
[107] First extremepoint
FilteringAlgorithm(EPEFA)
Extreme pointsextracting
Thresholds Binary BCH HashfunctionMD5
Verify overaudio andRF
AES
[105] Bumping bothdevices
Low-passFIR filte
10 statistics fea-tures
Decimal-to-binary
Binary
[109] Band-passfilter
5 statistics fea-tures
Thresholds Binary Hamming AES
[104] Thresholds Binary Hashed heuris-tic tree
[100] Onset ofdetected events
Coherence andquantized FFTcoefficien
Thresholds Binary Hashfunction
DH, CKP AES
[101] Peak of cross-correlation func-tion
Low-passfilte
Weight vector ofsegments, PCA
Pairwisenearestneighbor
Binary Hashfunction
[99] Tap two devices Decimal tobinary
Binary Diffie-Hellma
AES
[113] Low-passfilte
Peaks of signal -IPIs
Uniformquantizerwith qlevels
Gray Fuzzy commit-ment and BCH
[114] MadgwickandChebyshevbandpassfilte
Gait cycle detec-tion
Threshold Binary BCH Hashfunction
PasswordAuthenti-cated KeyExchanges(PAKE),fuzzycryptography
SHA-256
[115] First heel-strikeevent
Low-pass fil-ter
Thresholds Binary Index exchange Bit-wiseXORfunction
MAC AES
[116] First heel-strikeevent
Low-pass fil-ter
M-aryquantization
Binary Reed-Solomon(RS)
Bit-wiseXORfunction
MAC AES
[117] First heel-strikeevent
Low-pass fil-ter
Thresholds Binary Index exchange Bit-wiseXORfunction
MAC,HMAC-MD5
AES
[131] Discardingfour bits ofIPI
Inter-pulse inter-val (IPI)
Decimal togary code
Gray Hashfunction
NeymanPearsonLemma
AES-128,RSA
[123] Time-slottedchannelhopping-based(TSCH)
Savitzky-Golay(SG) filter,discardingthe least 3bits of IPI
Inter-pulse inter-val (IPI)
Decimal togary code
Gray CompressiveSensing (CS)
[138] High-pass filte,notch filter,ChebyshevIIR filter
Rectified EMGsignal
Fastdynamictimewarpingwith threelevels
Binary Binary GolayCode
the system’s restricted area. In a replay attack, the intrudersteals a packet from the network and forwards that packetto a service or application as if the intruder were the userwho originally sent the packet. Denial-of-service (DoS) anddistributed denial-of-service (DDoS) attacks are also examplesof active attacks, both of which work by preventing authorizedusers from accessing a specific resource on a network or theinternet (for example, flooding a device with more traffic thanit can handle).
Unlike a passive attack, an active attack is more likelyto be discovered quickly by the target upon executing it.For instance, we can stop the attacker from impersonatingthe nodes by using authentication mechanisms and intrusiondetection. To defend against reply attack, nonces and timetoken can be used to introduce fresh data [172]. Authenticationand anti-replay protection are the solutions suggested foravoiding denial of service [173], [174].
VI. LIMITATIONS AND CHALLENGES
The wearable device pairing methods should be lightweightwith fast computation, low storage, and low transmissionoverhead. Otherwise, the power and storage space of thebody sensors could be drained quickly. There are differentlimitations and challenges when trying to pair devices usingsensors’ data. The main limitations of the current approachesare that they require the user to do a specific action (e.g.,walking, handshake, gesture), and some devices should beplaced on a certain part of the body (e.g., wrist, arm, head) tocollect the desired signal. Also, some techniques need specialsensors like ECG and EMG sensors which are not used incommon wearable devices. These requirements can limit theusability of the proposed techniques. On the other hand, someother challenges can affect the accuracy of proposed methods;user’s motion artifact and health condition can challenge thepairing’s success rate. In some pairing methods, users mustbe static during the data collection phase. Furthermore, arecent study [175] revealed gait-based pairing approaches arevulnerable to video attack.
VII. CONCLUSION
By the rapid growth of smart wearable devices, pairing hasbecome the first and the most important concern in establishinga secure communication. Increasing various type of sensorsavailable in these devices have enabled autonomous pairingby exploiting common features of signals based on user’s ac-tivities and biometrics such as heartbeat, respiration, gait, andmotion. Accordingly, the number of research articles relatedto wearable device pairing has been increasing exponentially.
In this article, we have surveyed recent approaches inwearable device pairing and classified them according to thesignal types they have been used. We also descirebed thecommon required steps needed in most approaches to generatea common key in different devices. We then compared thefeatures and tools that authors used in each step. In addition,we mentioned the adversary models and countermeasures.Finally, we discussed challenges and limitations in pairing
wearable devices. This survey is expected to be useful forthe enhancement of spontaneous wearable device pairing.
REFERENCES
[1] K. Guk, G. Han, J. Lim, K. Jeong, T. Kang, E. K. Lim, and J. Jung,“Evolution of wearable devices with real-time disease monitoring forpersonalized healthcare,” Nanomaterials, vol. 9, no. 6, pp. 1–23, 2019.
[2] S. Vhaduri and C. Poellabauer, “Multi-modal biometric-based implicitauthentication of wearable device users,” IEEE Transactions on Infor-mation Forensics and Security, vol. 14, no. 12, pp. 3116–3125, 2019.
[3] M. K. Chong, R. Mayrhofer, and H. Gellersen, “A survey of user inter-action for spontaneous device association,” ACM Computing Surveys(CSUR), vol. 47, no. 1, pp. 1–40, 2014.
[4] S. Seneviratne, Y. Hu, T. Nguyen, G. Lan, S. Khalifa, K. Thilakarathna,M. Hassan, and A. Seneviratne, “A Survey of Wearable Devices andChallenges,” IEEE Communications Surveys and Tutorials, vol. 19,no. 4, pp. 2573–2620, 2017.
[5] A. Bianchi and I. Oakley, “Wearable authentication: Trends and oppor-tunities,” it - Information Technology, vol. 58, no. 5, 2016.
[6] F. John Dian, R. Vahidnia, and A. Rahmati, “Wearables and theInternet of Things (IoT), Applications, Opportunities, and Challenges:A Survey,” IEEE Access, vol. 8, pp. 69 200–69 211, 2020.
[7] T. Poongodi, R. Krishnamurthi, R. Indrakumari, P. Suresh, andB. Balusamy, Wearable devices and IoT. Springer InternationalPublishing, 2020, vol. 165. [Online]. Available: http://dx.doi.org/10.1007/978-3-030-23983-1 10
[8] A. Pantelopoulos and N. G. Bourbakis, “A survey on wearable sensor-based systems for health monitoring and prognosis,” IEEE Transactionson Systems, Man and Cybernetics Part C: Applications and Reviews,vol. 40, no. 1, pp. 1–12, 2010.
[9] S. Jayanth, M. Poorvi, R. Shreyas, B. Padmaja, and M. Sunil, “Wear-able device to measure heart beat using iot,” in 2017 InternationalConference on Inventive Systems and Control (ICISC). IEEE, 2017,pp. 1–5.
[10] A. Majumder, Y. A. ElSaadany, R. Young, and D. R. Ucci, “An energyefficient wearable smart iot system to predict cardiac arrest,” Advancesin Human-Computer Interaction, vol. 2019, 2019.
[11] A. Brezulianu, O. Geman, M. D. Zbancioc, M. Hagan, C. Aghion, D. J.Hemanth, L. H. Son et al., “Iot based heart activity monitoring usinginductive sensors,” Sensors, vol. 19, no. 15, p. 3284, 2019.
[12] S. Milici, J. Lorenzo, A. Lazaro, R. Villarino, and D. Girbau, “Wirelessbreathing sensor based on wearable modulated frequency selectivesurface,” IEEE Sensors Journal, vol. 17, no. 5, pp. 1285–1292, 2016.
[13] S. T. U. Shah, F. Badshah, F. Dad, N. Amin, and M. A. Jan, “Cloud-assisted iot-based smart respiratory monitoring system for asthmapatients,” in Applications of Intelligent Technologies in Healthcare.Springer, 2019, pp. 77–86.
[14] I. Mahbub, S. A. Pullano, H. Wang, S. K. Islam, A. S. Fiorillo, G. To,and M. Mahfouz, “A low-power wireless piezoelectric sensor-basedrespiration monitoring system realized in cmos process,” IEEE SensorsJournal, vol. 17, no. 6, pp. 1858–1864, 2017.
[15] D. Naranjo-Hernandez, A. Talaminos-Barroso, J. Reina-Tosina, L. M.Roa, G. Barbarov-Rostan, P. Cejudo-Ramos, E. Marquez-Martın, andF. Ortega-Ruiz, “Smart vest for respiratory rate monitoring of copdpatients based on non-contact capacitive sensing,” Sensors, vol. 18,no. 7, p. 2144, 2018.
[16] J. Wan, M. A. Al-awlaqi, M. Li, M. O’Grady, X. Gu, J. Wang, andN. Cao, “Wearable iot enabled real-time health monitoring system,”EURASIP Journal on Wireless Communications and Networking, vol.2018, no. 1, pp. 1–10, 2018.
[17] S. Yoshida, H. Miyaguchi, and T. Nakamura, “Development of tablet-shaped ingestible core-body thermometer powered by gastric acidbattery,” IEEE Sensors Journal, vol. 18, no. 23, pp. 9755–9762, 2018.
[18] F. Lamonaca, E. Balestrieri, I. Tudosa, F. Picariello, D. L. Carnı,C. Scuro, F. Bonavolonta, V. Spagnuolo, G. Grimaldi, and A. Colaprico,“An overview on internet of medical things in blood pressure monitor-ing,” in 2019 IEEE International Symposium on Medical Measurementsand Applications (MeMeA). IEEE, 2019, pp. 1–6.
[19] D. Murali, D. R. Rao, S. R. Rao, and M. Ananda, “Pulse oximetryand iot based cardiac monitoring integrated alert system,” in 2018international conference on advances in computing, communicationsand informatics (ICACCI). IEEE, 2018, pp. 2237–2243.
[20] B. Sargunam and S. Anusha, “Iot based mobile medical applicationfor smart insulin regulation,” in 2019 IEEE International Conferenceon Electrical, Computer and Communication Technologies (ICECCT).IEEE, 2019, pp. 1–5.
[21] C. Nave and O. Postolache, “Smart walker based iot physical reha-bilitation system,” in 2018 International Symposium in Sensing andInstrumentation in IoT Era (ISSI). IEEE, 2018, pp. 1–6.
[22] G. Yang, J. Deng, G. Pang, H. Zhang, J. Li, B. Deng, Z. Pang,J. Xu, M. Jiang, P. Liljeberg et al., “An iot-enabled stroke rehabilitationsystem based on smart wearable armband and machine learning,” IEEEjournal of translational engineering in health and medicine, vol. 6, pp.1–10, 2018.
[23] M. O. Agyeman and A. Al-Mahmood, “Design and implementationof a wearable device for motivating patients with upper and/or lowerlimb disability via gaming and home rehabilitation,” in 2019 FourthInternational Conference on Fog and Mobile Edge Computing (FMEC).IEEE, 2019, pp. 247–252.
[24] J. Qi, P. Yang, M. Hanneghan, S. Tang, and B. Zhou, “A hybridhierarchical framework for gym physical activity recognition andmeasurement using wearable sensors,” IEEE Internet of Things Journal,vol. 6, no. 2, pp. 1384–1393, 2018.
[25] D. Castro, W. Coral, C. Rodriguez, J. Cabra, and J. Colorado,“Wearable-based human activity recognition using an iot approach,”Journal of Sensor and Actuator Networks, vol. 6, no. 4, p. 28, 2017.
[26] H. Huang, X. Li, S. Liu, S. Hu, and Y. Sun, “Tribomotion: a self-powered triboelectric motion sensor in wearable internet of things forhuman activity recognition and energy harvesting,” IEEE Internet ofThings Journal, vol. 5, no. 6, pp. 4441–4453, 2018.
[27] W. Lu, F. Fan, J. Chu, P. Jing, and S. Yuting, “Wearable computingfor internet of things: A discriminant approach for human activityrecognition,” IEEE Internet of Things Journal, vol. 6, no. 2, pp. 2749–2759, 2018.
[28] L. Atallah, B. Lo, R. King, and G.-Z. Yang, “Sensor positioning foractivity recognition using wearable accelerometers,” IEEE transactionson biomedical circuits and systems, vol. 5, no. 4, pp. 320–329, 2011.
[29] M. Mathie, B. G. Celler, N. H. Lovell, and A. Coster, “Classificationof basic daily movements using a triaxial accelerometer,” Medical andBiological Engineering and Computing, vol. 42, no. 5, pp. 679–687,2004.
[30] A. Mannini and A. M. Sabatini, “Machine learning methods forclassifying human physical activity from on-body accelerometers,”Sensors, vol. 10, no. 2, pp. 1154–1175, 2010.
[31] E. Mencarini, A. Rapp, L. Tirabeni, and M. Zancanaro, “Designingwearable systems for sports: a review of trends and opportunities inhuman–computer interaction,” IEEE Transactions on Human-MachineSystems, vol. 49, no. 4, pp. 314–325, 2019.
[32] Z. Wang, M. Guo, and C. Zhao, “Badminton stroke recognition basedon body sensor networks,” IEEE Transactions on Human-MachineSystems, vol. 46, no. 5, pp. 769–775, 2016.
[33] A. Raina, T. Lakshmi, and S. Murthy, “Combat: Wearable technologybased training system for novice badminton players,” in 2017 IEEE17th International Conference on Advanced Learning Technologies(ICALT). IEEE, 2017, pp. 153–157.
[34] J. C. Maglott, J. Xu, and P. B. Shull, “Differences in arm motion timingcharacteristics for basketball free throw and jump shooting via a body-worn sensorized sleeve,” in 2017 IEEE 14th International Conferenceon Wearable and Implantable Body Sensor Networks (BSN). IEEE,2017, pp. 31–34.
[35] S. Bogers, C. Megens, and S. Vos, “Design for balanced engagement inmixed level sports teams,” in Proceedings of the 2017 CHI ConferenceExtended Abstracts on Human Factors in Computing Systems, 2017,pp. 994–1002.
[36] Z. Wang, H. Zhao, S. Qiu, and Q. Gao, “Stance-phase detection forzupt-aided foot-mounted pedestrian navigation system,” IEEE/ASMETransactions On Mechatronics, vol. 20, no. 6, pp. 3170–3181, 2015.
[37] Z. Wang, J. Wang, H. Zhao, N. Yang, and G. Fortino, “Canoesense:Monitoring canoe sprint motion using wearable sensors,” in 2016 IEEEInternational Conference on Systems, Man, and Cybernetics (SMC).IEEE, 2016, pp. 000 644–000 649.
[38] J. Pansiot, B. Lo, and G.-Z. Yang, “Swimming stroke kinematicanalysis with bsn,” in 2010 International Conference on Body SensorNetworks. IEEE, 2010, pp. 153–158.
[39] M. Bachlin, K. Forster, and G. Troster, “Swimmaster: a wearable assis-tant for swimmer,” in Proceedings of the 11th international conferenceon Ubiquitous computing, 2009, pp. 215–224.
[40] J. Hakkila, M. Alhonsuo, L. Virtanen, J. Rantakari, A. Colley, andT. Koivumaki, “Mydata approach for personal health–a service designcase for young athletes,” in 2016 49th Hawaii International Conferenceon System Sciences (HICSS). IEEE, 2016, pp. 3493–3502.
[41] J. Haladjian, M. Reif, and B. Brugge, “Vihapp: A wearable systemto support blind skiing,” in Proceedings of the 2017 ACM Interna-tional Joint Conference on Pervasive and Ubiquitous Computing andProceedings of the 2017 ACM International Symposium on WearableComputers, 2017, pp. 1033–1037.
[42] E. Niforatos, A. Fedosov, I. Elhart, and M. Langheinrich, “Augment-ing skiers’ peripheral perception,” in Proceedings of the 2017 ACMInternational Symposium on Wearable Computers, 2017, pp. 114–121.
[43] E. H. Chi, J. Song, and G. Corbin, “” killer app” of wearablecomputing: wireless force sensing body protectors for martial arts,”in Proceedings of the 17th annual ACM symposium on User interfacesoftware and technology, 2004, pp. 277–285.
[44] M. Pan, S. Salvi, and E. Brady, “Designing auditory feedback fromwearable weightlifting devices,” in Extended Abstracts of the 2018 CHIConference on Human Factors in Computing Systems, 2018, pp. 1–6.
[45] H. Havlucu, I. Bostan, A. Coskun, and O. Ozcan, “Understanding thelonesome tennis players: Insights for future wearables,” in Proceedingsof the 2017 CHI Conference Extended Abstracts on Human Factors inComputing Systems, 2017, pp. 1678–1685.
[46] M. Lapinski, E. Berkson, T. Gill, M. Reinold, and J. A. Paradiso, “Adistributed wearable, wireless sensor system for evaluating professionalbaseball pitchers and batters,” in 2009 International Symposium onWearable Computers. IEEE, 2009, pp. 131–138.
[47] H. Ghasemzadeh, V. Loseu, E. Guenterberg, and R. Jafari, “Sport train-ing using body sensor networks: A statistical approach to measure wristrotation for golf swing,” in Proceedings of the Fourth InternationalConference on Body Area Networks, 2009, pp. 1–8.
[48] S. Seneviratne, Y. Hu, T. Nguyen, G. Lan, S. Khalifa, K. Thilakarathna,M. Hassan, and A. Seneviratne, “A survey of wearable devices andchallenges,” IEEE Communications Surveys & Tutorials, vol. 19, no. 4,pp. 2573–2620, 2017.
[49] C. Wang, J. Liu, X. Guo, Y. Wang, and Y. Chen, “Wristspy: Snoopingpasscodes in mobile payment using wrist-worn wearables,” in IEEEINFOCOM 2019-IEEE Conference on Computer Communications.IEEE, 2019, pp. 2071–2079.
[50] T. Nguyen and N. D. Memon, “Smartwatches locking methods: Acomparative study.” in SOUPS, 2017.
[51] R. Kumar, V. V. Phoha, and R. Raina, “Authenticating users throughtheir arm movement patterns,” arXiv preprint arXiv:1603.02211, 2016.
[52] R. C. Shit, S. Sharma, D. Puthal, and A. Y. Zomaya, “Location ofthings (lot): A review and taxonomy of sensors localization in iotinfrastructure,” IEEE Communications Surveys & Tutorials, vol. 20,no. 3, pp. 2028–2061, 2018.
[53] Y. Kim, H. Shin, and H. Cha, “Smartphone-based wi-fi pedestrian-tracking system tolerating the rss variance problem,” in 2012 IEEE In-ternational Conference on Pervasive Computing and Communications.IEEE, 2012, pp. 11–19.
[54] W. Saadeh, S. A. Butt, and M. A. B. Altaf, “A patient-specific singlesensor iot-based wearable fall prediction and detection system,” IEEEtransactions on neural systems and rehabilitation engineering, vol. 27,no. 5, pp. 995–1003, 2019.
[55] W. Saadeh, M. A. B. Altaf, and S. A. Butt, “A wearable neuro-degenerative diseases detection system based on gait dynamics,” in2017 IFIP/IEEE International Conference on Very Large Scale Inte-gration (VLSI-SoC). IEEE, 2017, pp. 1–6.
[56] L. Ren and Y. Peng, “Research of fall detection and fall preventiontechnologies: A systematic review,” IEEE Access, vol. 7, pp. 77 702–77 722, 2019.
[57] N. Otanasap, “Pre-impact fall detection based on wearable device usingdynamic threshold model,” in 2016 17th International Conference onParallel and Distributed Computing, Applications and Technologies(PDCAT). IEEE, 2016, pp. 362–365.
[58] M. Nyan, F. E. Tay, and E. Murugasu, “A wearable system for pre-impact fall detection,” Journal of biomechanics, vol. 41, no. 16, pp.3475–3481, 2008.
[59] P. Choudhary, R. Sharma, G. Singh, S. Das et al., “A survey paperon drowsiness detection & alarm system for drivers,” International
Research Journal of Engineering and Technology (IRJET), vol. 3,no. 12, pp. 1433–1437, 2016.
[60] W.-J. Chang, L.-B. Chen, and Y.-Z. Chiou, “Design and implementationof a drowsiness-fatigue-detection system based on wearable smartglasses to increase road safety,” IEEE Transactions on ConsumerElectronics, vol. 64, no. 4, pp. 461–469, 2018.
[61] S. R. Dhole, A. Kashyap, A. N. Dangwal, and R. Mohan, “A novelhelmet design and implementation for drowsiness and fall detectionof workers on-site using eeg and random-forest classifier,” ProcediaComputer Science, vol. 151, pp. 947–952, 2019.
[62] F. Wu, J.-M. Redoute, and M. R. Yuce, “A self-powered wearablebody sensor network system for safety applications,” in 2018 IEEESENSORS. IEEE, 2018, pp. 1–4.
[63] M. Serbanescu, V. Placinta, O. Hutanu, and C. Ravariu, “Smart, lowpower, wearable multi-sensor data acquisition system for environmentalmonitoring,” in 2017 10th International Symposium on AdvancedTopics in Electrical Engineering (ATEE). IEEE, 2017, pp. 118–123.
[64] A. G. Perez, D. Lobo, F. Chinello, G. Cirio, M. Malvezzi, J. San Martin,D. Prattichizzo, and M. A. Otaduy, “Optimization-based wearabletactile rendering,” IEEE transactions on haptics, vol. 10, no. 2, pp.254–264, 2016.
[65] M. Maisto, C. Pacchierotti, F. Chinello, G. Salvietti, A. De Luca, andD. Prattichizzo, “Evaluation of wearable haptic systems for the fingersin augmented reality applications,” IEEE transactions on haptics,vol. 10, no. 4, pp. 511–522, 2017.
[66] M. Fomichev, F. Alvarez, D. Steinmetzer, P. Gardner-Stephen, andM. Hollick, “Survey and systematization of secure device pairing,”arXiv, vol. 20, no. 1, pp. 517–550, 2017.
[67] J. Unar, W. C. Seng, and A. Abbasi, “A review of biometric technologyalong with trends and prospects,” Pattern recognition, vol. 47, no. 8,pp. 2673–2688, 2014.
[68] Y. Zeng, A. Pande, J. Zhu, and P. Mohapatra, “Wearia: Wearable deviceimplicit authentication based on activity information,” in 2017 IEEE18th International Symposium on A World of Wireless, Mobile andMultimedia Networks (WoWMoM). IEEE, 2017, pp. 1–9.
[69] K.-A. Shim, “A survey of public-key cryptographic primitives inwireless sensor networks,” IEEE Communications Surveys & Tutorials,vol. 18, no. 1, pp. 577–601, 2015.
[70] N. D. Sarier, “A survey of distributed biometric authentication sys-tems,” BIOSIG 2009: biometrics and electronic signatures, 2009.
[71] R. Cavallari, F. Martelli, R. Rosini, C. Buratti, and R. Verdone, “Asurvey on wireless body area networks: Technologies and designchallenges,” IEEE Communications Surveys & Tutorials, vol. 16, no. 3,pp. 1635–1657, 2014.
[72] S. Movassaghi, M. Abolhasan, J. Lipman, D. Smith, and A. Jamalipour,“Wireless body area networks: A survey,” IEEE Communicationssurveys & tutorials, vol. 16, no. 3, pp. 1658–1686, 2014.
[73] S. S. Javadi and M. A. Razzaque, “Security and Privacy in WirelessBody Area Networks for Health Care Applications,” pp. 165–187,2013.
[74] M. Masdari and S. Ahmadzadeh, “Comprehensive analysis of theauthentication methods in wireless body area networks,” Security andcommunication networks, vol. 9, no. 17, pp. 4777–4803, 2016.
[75] V. Mainanwal, M. Gupta, and S. K. Upadhayay, “A survey on wirelessbody area network: Security technology and its design methodologyissue,” ICIIECS 2015 - 2015 IEEE International Conference on Inno-vations in Information, Embedded and Communication Systems, no. I,pp. 1–5, 2015.
[76] M. R. K. Naik and P. Samundiswary, “Wireless body area networksecurity issues - Survey,” 2016 International Conference on ControlInstrumentation Communication and Computational Technologies, IC-CICCT 2016, pp. 190–194, 2017.
[77] S. Bhatt and T. Santhanam, “Keystroke dynamics for biometricauthentication-A survey,” Proceedings of the 2013 International Con-ference on Pattern Recognition, Informatics and Mobile Engineering,PRIME 2013, pp. 17–23, 2013.
[78] M. Karnan, M. Akila, and N. Krishnaraj, “Biometric personalauthentication using keystroke dynamics: A review,” Applied SoftComputing Journal, vol. 11, no. 2, pp. 1565–1573, 2011. [Online].Available: http://dx.doi.org/10.1016/j.asoc.2010.08.003
[79] M. Abuhamad, A. Abusnaina, D. Nyang, and D. Mohaisen, “Sensor-Based Continuous Authentication of Smartphones’ Users Using Behav-ioral Biometrics: A Contemporary Survey,” IEEE Internet of ThingsJournal, vol. 8, no. 1, pp. 65–84, 2021.
[80] S. Al-Janabi, I. Al-Shourbaji, M. Shojafar, and S. Shamshirband,“Survey of main challenges (security and privacy) in wireless bodyarea networks for healthcare applications,” Egyptian InformaticsJournal, vol. 18, no. 2, pp. 113–122, 2017. [Online]. Available:http://dx.doi.org/10.1016/j.eij.2016.11.001
[81] S. Zou, Y. Xu, H. Wang, Z. Li, S. Chen, and B. Hu, “A Survey onSecure Wireless Body Area Networks,” Security and CommunicationNetworks, vol. 2017, 2017.
[82] B. Narwal and A. K. Mohapatra, “A review on authentication protocolsin wireless body area networks (WBAN),” Proceedings of the 3rd In-ternational Conference on Contemporary Computing and Informatics,IC3I 2018, pp. 227–232, 2018.
[83] M. Usman, M. R. Asghar, I. S. Ansari, and M. Qaraqe, “Security inwireless body area networks: From in-body to off-body communica-tions,” IEEE Access, vol. 6, pp. 58 064–58 074, 2018.
[84] M. S. Arshad Malik, M. Ahmed, T. Abdullah, N. Kousar, M. N.Shumaila, and M. Awais, “Wireless body area network security andprivacy issue in E-healthcare,” International Journal of AdvancedComputer Science and Applications, vol. 9, no. 4, pp. 209–215, 2018.
[85] L. V. Morales, D. Delgado-Ruiz, and S. J. Rueda, “Comprehensivesecurity for body area networks: A survey,” Int. J. Netw. Secur, vol. 21,pp. 342–354, 2019.
[86] M. Kompara and M. Holbl, “Survey on security in intra-body areanetwork communication,” Ad Hoc Networks, vol. 70, pp. 23–43, 2018.[Online]. Available: https://doi.org/10.1016/j.adhoc.2017.11.006
[87] R. Nidhya and S. Karthik, “Security and privacy issues in remotehealthcare systems using wireless body area networks,” in Body AreaNetwork Challenges and Solutions. Springer, 2019, pp. 37–53.
[88] A. Joshi and A. K. Mohapatra, “Authentication protocols for wirelessbody area network with key management approach,” Journal of Dis-crete Mathematical Sciences and Cryptography, vol. 22, no. 2, pp.219–240, 2019.
[89] S. Chaudhary, A. Singh, and K. Chatterjee, “Wireless Body SensorNetwork (WBSN) Security and Privacy Issues: A Survey,” pp. 515–521, 2019.
[90] M. Hussain, A. Mehmood, S. Khan, M. A. Khan, and Z. Iqbal,“Authentication Techniques and Methodologies used in Wireless BodyArea Networks,” Journal of Systems Architecture, vol. 101, p. 101655,2019. [Online]. Available: https://doi.org/10.1016/j.sysarc.2019.101655
[91] M. Roy, C. Chowdhury, and N. Aslam, “Security and privacy issuesin wireless sensor and body area networks,” in Handbook of computernetworks and cyber security. Springer, 2020, pp. 173–200.
[92] B. Narwal and A. K. Mohapatra, A survey on security andauthentication in wireless body area networks. Elsevier B.V., 2021,vol. 113. [Online]. Available: https://doi.org/10.1016/j.sysarc.2020.101883
[93] A. Al Abdulwahid, N. Clarke, I. Stengel, S. Furnell, and C. Reich,“Continuous and transparent multimodal authentication: reviewing thestate of the art,” Cluster Computing, vol. 19, no. 1, pp. 455–474, 2016.
[94] R. Spolaor, Q. Q. Li, M. Monaro, M. Conti, L. Gamberini, andG. Sartori, “Biometric authentication methods on smartphones: Asurvey,” PsychNology Journal, vol. 14, no. 2-3, pp. 87–98, 2016.
[95] T. Neal and D. Woodard, “Surveying Biometric Authentication forMobile Device Security,” Journal of Pattern Recognition Research,vol. 11, no. 1, pp. 74–110, 2016.
[96] R. Mayrhofer, J. Fuß, and I. Ion, “Uacap: A unified auxiliary channelauthentication protocol,” IEEE Transactions on Mobile Computing,vol. 12, no. 4, pp. 710–721, 2012.
[97] R. Ferrero, F. Gandino, B. Montrucchio, M. Rebaudengo, A. Velasco,and I. Benkhelifa, “On gait recognition with smartphone accelerome-ter,” in 2015 4th Mediterranean Conference on Embedded Computing(MECO). IEEE, 2015, pp. 368–373.
[98] A. L. Fantana, S. Ramachandran, C. H. Schunck, and M. Talamo,“Movement based biometric authentication with smartphones,” in 2015International Carnahan Conference on Security Technology (ICCST).IEEE, 2015, pp. 235–239.
[99] R. Jin, L. Shi, K. Zeng, A. Pande, and P. Mohapatra, “Magpairing:Pairing smartphones in close proximity using magnetometers,” IEEETransactions on Information Forensics and Security, vol. 11, no. 6, pp.1306–1320, 2015.
[100] A. LaMarca, M. Langheinrich, and K. N. Truong, “PervasiveComputing, 5th International Conference, PERVASIVE 2007, Toronto,Canada, May 13-16, 2007, Proceedings.” vol. 4480, no. May
2007, 2007. [Online]. Available: http://link.springer.com/10.1007/978-3-540-72037-9%5Cn
[101] D. Bichler, G. Stromberg, M. Huemer, and M. Low, “Key generationbased on acceleration data of shaking processes,” Lecture Notes inComputer Science (including subseries Lecture Notes in ArtificialIntelligence and Lecture Notes in Bioinformatics), vol. 4717 LNCS,pp. 304–317, 2007.
[102] Y. Shen, F. Yang, B. Du, W. Xu, C. Luo, and H. Wen, “Shake-n-shack: Enabling secure data exchange between smart wearables viahandshakes,” arXiv, pp. 1–10, 2018.
[103] Q. Jiang, X. Huang, N. Zhang, K. Zhang, X. Ma, and J. Ma, “Shake tocommunicate: Secure handshake acceleration-based pairing mechanismfor wrist worn devices,” IEEE Internet of Things Journal, vol. 6, no. 3,pp. 5618–5630, 2019.
[104] B. Groza and R. Mayrhofer, “SAPHE: Simple accelerometer basedwireless pairing with heuristic trees,” ACM International ConferenceProceeding Series, no. 2, pp. 161–168, 2012.
[105] H. Yuzuguzel, J. Niemi, S. Kiranyaz, M. Gabbouj, and T. Heinz,“ShakeMe: Key generation from shared motion,” Proceedings - 15thIEEE International Conference on Computer and Information Tech-nology, CIT 2015, 14th IEEE International Conference on UbiquitousComputing and Communications, IUCC 2015, 13th IEEE InternationalConference on Dependable, Autonomic and Se, pp. 2130–2133, 2015.
[106] B. Groza, A. Berdich, C. Jichici, and R. Mayrhofer, “SecureAccelerometer-based Pairing of Mobile Devices in Multi-modal Trans-port,” IEEE Access, vol. 8, pp. 1–1, 2020.
[107] M. Cao, L. Wang, H. Xu, D. Chen, C. Lou, N. Zhang, Y. Zhu, andZ. Qin, “Sec-D2D: A Secure and Lightweight D2D CommunicationSystem with Multiple Sensors,” IEEE Access, vol. 7, pp. 33 759–33 770,2019.
[108] E. Bejder, A. K. Mathiasen, M. De Donno, N. Dragoni, and X. Fafoutis,“SHAKE: SHared Acceleration Key Establishment for Resource-Constrained IoT Devices,” IEEE World Forum on Internet of Things,WF-IoT 2020 - Symposium Proceedings, 2020.
[109] S. Sen and D. Kotz, “Vibering: Using vibrations from a smart ring asan out-of-band channel for sharing secret keys,” in Proceedings of the10th International Conference on the Internet of Things, 2020, pp. 1–8.
[110] M. O. Derawi, C. Nickel, P. Bours, and C. Busch, “Unobtrusive user-authentication on mobile phones using biometric gait recognition,” in2010 Sixth International Conference on Intelligent Information Hidingand Multimedia Signal Processing. IEEE, 2010, pp. 306–311.
[111] Y. Zhong and Y. Deng, “Sensor orientation invariant mobile gaitbiometrics,” in IEEE international joint conference on biometrics.IEEE, 2014, pp. 1–8.
[112] J. Lester, B. Hannaford, and G. Borriello, “”Are you with me?”- Using accelerometers to determine if two devices are carried bythe same person,” Lecture Notes in Computer Science (includingsubseries Lecture Notes in Artificial Intelligence and Lecture Notesin Bioinformatics), vol. 3001, pp. 33–50, 2004.
[113] Y. Sun, C. Wong, G. Z. Yang, and B. Lo, “Secure key generation usinggait features for Body Sensor Networks,” 2017 IEEE 14th InternationalConference on Wearable and Implantable Body Sensor Networks, BSN2017, pp. 206–210, 2017.
[114] D. Schurmann, A. Brusch, N. Nguyen, S. Sigg, and L. Wolf, “Moveslike Jagger: Exploiting variations in instantaneous gait for spontaneousdevice pairing,” Pervasive and Mobile Computing, vol. 47, no. May2017, pp. 1–12, 2018.
[115] W. Xu, G. Revadigar, C. Luo, N. Bergmann, and W. Hu, “Walkie-talkie:Motion-assisted automatic key generation for secure on-body devicecommunication,” in 2016 15th ACM/IEEE International Conferenceon Information Processing in Sensor Networks (IPSN). IEEE, 2016,pp. 1–12.
[116] W. Xu, C. Javali, G. Revadigar, C. Luo, N. Bergmann, and W. Hu,“Gait-key: A gait-based shared secret key generation protocol forwearable devices,” ACM Transactions on Sensor Networks (TOSN),vol. 13, no. 1, pp. 1–27, 2017.
[117] W. Xu and G. Lan, “Gait-based smart pairing system for personalwearable devices,” in Medical Internet of Things (m-IoT)-EnablingTechnologies and Emerging Applications. IntechOpen, 2019.
[118] A. Brusch, N. Nguyen, D. Schurmann, S. Sigg, and L. Wolf, “SecurityProperties of Gait for Mobile Device Pairing,” IEEE Transactions onMobile Computing, vol. 19, no. 3, pp. 697–710, 2020.
[119] K. M. S. Thotahewa, J. M. Redoute, and M. R. Yuce, Ultra widebandwireless body area networks, 2014, vol. 9783319052878.
[120] S. Sujatha and R. Govindaraju, “A Secure Crypto based ECG DataCommunication using Modified SPHIT and Modified Quasigroup En-cryption,” International Journal of Computer Applications, vol. 78,no. 6, pp. 27–33, 2013.
[121] W. Wang, H. Wang, M. Hempel, D. Peng, H. Sharif, and H. H. Chen,“Secure stochastic ECG signals based on gaussian mixture model for e-healthcare systems,” IEEE Systems Journal, vol. 5, no. 4, pp. 564–573,2011.
[122] F. Rundo, S. Conoci, A. Ortis, and S. Battiato, “An advanced bio-inspired photoplethysmography (PPG) and ECG pattern recognitionsystem for medical assessment,” Sensors (Switzerland), vol. 18, no. 2,2018.
[123] Q. Lin, W. Xu, J. Liu, A. Khamis, W. Hu, M. Hassan, and A. Senevi-ratne, “H2B: Heartbeat-based secret key generation using piezo vibra-tion sensors,” arXiv, pp. 265–276, 2019.
[124] L. Wang, K. Huang, K. Sun, W. Wang, C. Tian, L. Xie, and Q. Gu,“Unlock with your heart: Heartbeat-based authentication on commer-cial mobile phones,” Proceedings of the ACM on interactive, mobile,wearable and ubiquitous technologies, vol. 2, no. 3, pp. 1–22, 2018.
[125] J. Ramos-Castro, J. Moreno, H. Miranda-Vidal, M. A. Garcıa-Gonzalez, M. Fernandez-Chimeno, G. Rodas, and L. Capdevila, “Heartrate variability analysis using a seismocardiogram signal,” in 2012annual international conference of the IEEE engineering in medicineand biology society. IEEE, 2012, pp. 5642–5645.
[126] W. B. A. Karaa, Biomedical image analysis and mining techniques forimproved health outcomes. IGI Global, 2015.
[127] F. Sufi, I. Khalil, and J. Hu, “ECG-Based Authentication,” Handbookof Information and Communication Security, pp. 309–331, 2010.
[128] E. Okoh, “Biometrics Solutions in e-Health Security: A ComprehensiveLiterature Review,” Spine, vol. 19, no. Supplement, pp. 2274S–2278S,2015.
[129] R. McCraty and F. Shaffer, “Heart rate variability: New perspectives onphysiological mechanisms, assessment of self-regulatory capacity, andhealth risk,” Global Advances In Health and Medicine, vol. 4, no. 1,pp. 46–61, 2015.
[130] S. Cherukuri, K. K. Venkatasubramanian, and S. K. Gupta, “Biosec:A biometric based approach for securing communication in wire-less networks of biosensors implanted in the human body,” in 2003International Conference on Parallel Processing Workshops, 2003.Proceedings. IEEE, 2003, pp. 432–439.
[131] M. Rostami and A. Juels, “Authentication for Implanted MedicalDevices Categories and Subject Descriptors,” Ccs, pp. 1099–1111,2013.
[132] P. A. Obrist, Cardiovascular psychophysiology: A perspective.Springer Science & Business Media, 2012.
[133] R. Merletti and P. J. Parker, Electromyography: physiology, engineer-ing, and non-invasive applications. John Wiley & Sons, 2004, vol. 11.
[134] S. R. Devasahayam, Signals and systems in biomedical engineering:signal processing and physiological systems modeling. SpringerScience & Business Media, 2012.
[135] “Myo armband,” https://www.myo.com.[136] “Athos gear,” https://www.liveathos.com.[137] “Leo smartband,” http://leohelps.com.[138] L. Yang, W. Wang, and Q. Zhang, “Secret from muscle: Enabling
secure pairing with electromyography,” Proceedings of the 14th ACMConference on Embedded Networked Sensor Systems, SenSys 2016, pp.28–41, 2016.
[139] K. M. Iftekharuddin and A. Awwal, “Sampling and Quantization,” FieldGuide to Image Processing, no. i, pp. 5–5, 2012.
[140] D. Schurmann and S. Sigg, “Secure communication based on ambientaudio,” IEEE Transactions on mobile computing, vol. 12, no. 2, pp.358–370, 2011.
[141] A. Elsts, S. Duquennoy, X. Fafoutis, G. Oikonomou, R. Piechocki,and I. Craddock, “Microsecond-accuracy time synchronization usingthe ieee 802.15. 4 tsch protocol,” in 2016 IEEE 41st conference onlocal computer networks workshops (LCN Workshops). IEEE, 2016,pp. 156–164.
[142] M. B. del Rosario, S. J. Redmond, and N. H. Lovell, “Tracking theevolution of smartphone sensing for monitoring human movement,”Sensors (Switzerland), vol. 15, no. 8, pp. 18 901–18 933, 2015.
[143] M. Naseer, M. A. Azam, M. Ehatisham-Ul-Haq, W. Ejaz, andA. Khalid, “ADLAuth: Passive authentication based on activity ofdaily living using heterogeneous sensing in smart cities,” Sensors(Switzerland), vol. 19, no. 11, 2019.
[144] Y.-P. Chen, J.-Y. Yang, S.-N. Liou, G.-Y. Lee, and J.-S. Wang, “Onlineclassifier construction algorithm for human activity detection using atri-axial accelerometer,” Applied Mathematics and Computation, vol.205, no. 2, pp. 849–860, 2008.
[145] Y. He and Y. Li, “Physical activity recognition utilizing the built-in kinematic sensors of a smartphone,” International Journal of Dis-tributed Sensor Networks, vol. 9, no. 4, p. 481580, 2013.
[146] U. Maurer, A. Rowe, A. Smailagic, and D. Siewiorek, “Location andactivity recognition using ewatch: A wearable sensor platform,” inAmbient intelligence in everyday life. Springer, 2006, pp. 86–102.
[147] J. Saunders, “Real-time discrimination of broadcast speech/music,” in1996 IEEE International Conference on Acoustics, Speech, and SignalProcessing Conference Proceedings, vol. 2. IEEE, 1996, pp. 993–996.
[148] C. V. Bouten, K. T. Koekkoek, M. Verduin, R. Kodde, and J. D.Janssen, “A triaxial accelerometer and portable data processing unitfor the assessment of daily physical activity,” IEEE transactions onbiomedical engineering, vol. 44, no. 3, pp. 136–147, 1997.
[149] J. Parkka, M. Ermes, P. Korpipaa, J. Mantyjarvi, J. Peltola, andI. Korhonen, “Activity classification using realistic data from wearablesensors,” IEEE Transactions on information technology in biomedicine,vol. 10, no. 1, pp. 119–128, 2006.
[150] A. Moncada-Torres, K. Leuenberger, R. Gonzenbach, A. Luft, andR. Gassert, “Activity classification based on inertial and barometricpressure sensors at different anatomical locations,” Physiological mea-surement, vol. 35, no. 7, p. 1245, 2014.
[151] M. Ehatisham-ul Haq, M. A. Azam, J. Loo, K. Shuang, S. Islam,U. Naeem, and Y. Amin, “Authentication of smartphone users basedon activity recognition and mobile sensing,” Sensors, vol. 17, no. 9, p.2043, 2017.
[152] T. Huynh and B. Schiele, “Analyzing features for activity recognition,”in Proceedings of the 2005 joint conference on Smart objects andambient intelligence: innovative context-aware services: usages andtechnologies, 2005, pp. 159–163.
[153] K. Sayood, Introduction to data compression. Morgan Kaufmann,2017.
[154] S. Y. Ho and D. J. Kleitman, “An odd kind of BCH code,” DiscreteApplied Mathematics, vol. 161, no. 9, pp. 1216–1220, 2013. [Online].Available: http://dx.doi.org/10.1016/j.dam.2012.08.021
[155] “Error correction code, wikipedia, the free encyclopedia,” https://en.wikipedia.org/wiki/Error correction code.
[156] R. C. Bose and D. K. Ray-Chaudhuri, “On a class of error correctingbinary group codes,” Information and control, vol. 3, no. 1, pp. 68–79,1960.
[157] I. S. Reed and G. Solomon, “Polynomial codes over certain finitefields,” Journal of the society for industrial and applied mathematics,vol. 8, no. 2, pp. 300–304, 1960.
[158] R. W. Hamming, “Error detecting and error correcting codes,” The Bellsystem technical journal, vol. 29, no. 2, pp. 147–160, 1950.
[159] M. J. Golay, “Notes on digital coding,” Proc. IEEE, vol. 37, p. 657,1949.
[160] “Bch code, wikipedia, the free encyclopedia,” https://en.wikipedia.org/wiki/BCH code.
[161] P. Shrivastava and U. P. Singh, “Error detection and correction usingreed solomon codes,” International Journal of Advanced Research inComputer Science and Software Engineering, vol. 3, no. 8, 2013.
[162] A. Sahai and B. Waters, “Fuzzy identity-based encryption,” in Annualinternational conference on the theory and applications of crypto-graphic techniques. Springer, 2005, pp. 457–473.
[163] A. Anees and Y.-P. P. Chen, “Discriminative binary feature learning andquantization in biometric key generation,” Pattern Recognition, vol. 77,pp. 289–305, 2018.
[164] D. L. Donoho, “Compressed sensing,” IEEE Transactions on Informa-tion Theory, vol. 52, no. 4, pp. 1289–1306, 2006.
[165] R. Chartrand, “Nonconvex compressed sensing and error correction,” in2007 IEEE International Conference on Acoustics, Speech and SignalProcessing - ICASSP ’07, vol. 3, 2007, pp. III–889–III–892.
[166] H. Zoerlein, M. Shehata, and M. Bossert, “Concatenated compressedsensing-based error correcting codes,” in SCC 2013; 9th InternationalITG Conference on Systems, Communication and Coding, 2013, pp.1–5.
[167] M. Hosseinzadeh, B. Vo, M. Y. Ghafour, and S. Naghipour,Electrocardiogram signals-based user authentication systems usingsoft computing techniques. Springer Netherlands, 2021, vol. 54, no. 1.[Online]. Available: https://doi.org/10.1007/s10462-020-09863-0
[168] A. Rukhin, J. Soto, J. Nechvatal, M. Smid, and E. Barker, “A statis-tical test suite for random and pseudorandom number generators forcryptographic applications,” Booz-allen and hamilton inc mclean va,Tech. Rep., 2001.
[169] Y. Yang, L. Wu, G. Yin, L. Li, and H. Zhao, “A survey on security andprivacy issues in internet-of-things,” IEEE Internet of Things Journal,vol. 4, no. 5, pp. 1250–1258, 2017.
[170] J. Deogirikar and A. Vidhate, “Security attacks in iot: A survey,” in2017 International Conference on I-SMAC (IoT in Social, Mobile,Analytics and Cloud)(I-SMAC). IEEE, 2017, pp. 32–37.
[171] J.-C. Kao and R. Marculescu, “Eavesdropping minimization via trans-mission power control in ad-hoc wireless networks,” in 2006 3rdAnnual IEEE Communications Society on Sensor and Ad Hoc Com-munications and Networks, vol. 2. IEEE, 2006, pp. 707–714.
[172] S. Malladi, J. Alves-Foss, and R. B. Heckendorn, “On preventing replayattacks on security protocols,” IDAHO UNIV MOSCOW DEPT OFCOMPUTER SCIENCE, Tech. Rep., 2002.
[173] Y. Wang, G. Attebury, and B. Ramamurthy, “A survey of security issuesin wireless sensor networks,” 2006.
[174] J. P. Walters, Z. Liang, W. Shi, and V. Chaudhary, “Wireless sensornetwork security: A survey,” Security in distributed, grid, mobile, andpervasive computing, vol. 1, no. 367, p. 6, 2007.
[175] A. Bruesch, N. Nguyen, D. Schurmann, S. Sigg, and L. Wolf, “Securityproperties of gait for mobile device pairing,” IEEE Transactions onMobile Computing, vol. 19, no. 3, pp. 697–710, 2019.
JAFAR POURBEMANY received the BSc degreein Electronics and the MSc degree in Communi-cations from Yazd University, in 2009 and 2014,respectively. He is currently pursuing the Ph.D. de-gree in Computer Science with the Network Securityand Privacy Research Laboratory, Cleveland StateUniversity, USA, where he is currently a ResearchAssistant. His research interests include networksecurity, body area networks, wireless sensor net-works, and machine learning.
YE ZHU received the BSc degree in 1994 fromShanghai JiaoTong University and the MSc degreein 2002 from Texas A& M University. He receivedthe PhD degree from the Electrical and ComputerEngineering Department at Texas A& M Univer-sity. He is currently an associate professor in theDepartment of Electrical and Computer Engineeringat Cleveland State University. His research interestsinclude network security, traffic engineering, andwireless sensor networks. He is a member of theIEEE and the IEEE Computer Society.
Riccardo Bettati received the diploma in informat-ics from the Swiss Federal Institute of Technology(ETH), Zuerich, Switzerland, in 1988 and the PhDdegree from the University of Illinois at Urbana-Champaign, in 1994. He is a professor with theDepartment of Computer Science, Texas A&M Uni-versity, where he has been leading the Real-TimeSystems Research Group and until 2015 the Centerfor Information Assurance and Security. From 1993to 1995, he held a postdoctoral position with theInternational Computer Science Institute in Berkeley
and with the University of California at Berkeley. His research interestsinclude traffic analysis and privacy, real-time distributed systems, real-timecommunication, and network support for resilient distributed applications. Hewas the program and general chairs of The IEEE Real-Time and EmbeddedTechnology and Applications Symposia in 2002 and 2003, respectively. Heshares Best Paper awards with collaborators and students in the IEEE NationalAerospace and Electronics Conference and in the Euromicro Conference onReal-Time Systems.
