A Trt Evalatio Framework for Seor Reaig i Boy Area Seor Network Vih Bi, Richar Verhoeve, Joha Lkkie, Rafal Kocielik Departmet of Mathematic a Compter Sciece Eihove Uiverity of Techology t.v.i, p.h.f.m.verhoeve, j.j.lkkie, r..kocielik}@te.l ABSTRACT Thi paper aree a framework to evalate trtworthie of a Boy Area Seor Network (BASN), i particlar, of eor reaig. We how that ch trtworthie i to e iterprete with repect to a certai tatemet or goal; it evalatio i ae o qality apect erive from oervatio a opiio from other. We examie relevat qality apect of eor reaig which correpo to potetial eviatig ehavior of eor. We the look at how to erive ch qalitie from oervatio takig certaity ito the evalatio a well a ecay over time. We evelop a exteio of jective logic for thi prpoe a we how how we ca compte qality propertie withot torig log time erie. We the emotrate thi for two example, iclig Galvaic Ski Repoe (GSR) a Electrocariography (ECG) ee ata.. INTRODUCTION We coier a moitorig a reaoig framework for evalatig trtworthie of collecte ata i a Boy Area Seor Network (BASN) ae o Sjective Logic [, 4]. A BASN coit of eor oe mote o or implate i the hma oy together with a more powerfl evice which we call the oy h. Thi h i capale of torig ata a rig applicatio-pecific compoet; it cotrol the BASN a act a a igle acce poit. I crret techology, the h i typically a mart phoe. A BASN i a ytem of a larger ytem; typically, it coect to a ack-e throgh the Iteret for the prpoe of ata exchage a (re)cofigratio. I ome eig there i little itelligece at the er ie of the etwork except for rigig itermittet coectivity. However, i or viio a BASN ca operate iepeetly er cotrol of it ower, iclig torig a maagig ata. Toay iitial exploratio of BASN are motly i the omai of elf-moitorig for well-eig a port achievemet th avoiig trict meical reglatio, thogh we ca alreay ee a evelopmet i which health moitorig, iitiate y either a meical pecialit or the er himelf, ecome part of the maagemet of er health. A importat qetio the i to what extet a takeholer (er, pecialit) ca rely o a BASN, i.e., how he etermie it trtworthie. Trtworthie, i geeral, refer to a relatio amog etitie, where oe relie o the other. At the er-ytem iteractio level, thi i commoly ae o a high traparecy a a clear iteractio, complemete with reliale a correct operatio. Whe e iie the ytem, trt a trtworthie refer to relatio etwee etitie i the ytem itelf. Trt a a trt moel form the ai of ytem eciio (e.g., whether or ot a operatig ytem accept a applicatio, how it aert the qality of otcome or ecie to take a actio). I crret ytem ch trt i motly implicit i ytem operatio like checkig a certificate or akig a er coet. I or work we chooe to relate the trt to the ject of the eciio. For example, oe may trt a imple ECG eor to eliver a heart rate ervice t ot a etaile ECG. Thi alo gget that i the evalatio of ch trt there are ojective elemet (fact like mearemet, or the opiio of aother etity o the ject matter) a jective elemet (the vale a trtor aig to ch a fact). Overall, trt i ilt from ome compoitio of propertie, typically etermie throgh moitorig, a i relate to the ervice that i expecte. Followig [3], trt i therefore efie a the egree to which a trtor ha a jtifiale elief that, i a give cotext, a trtee will live p to a give et of tatemet aot it ehavior. I [3], a trt maagemet moel wa preete alog with a compoet-ae architectre to maitai the BASN a a trtworthy platform er chage i applicatio ch a yamic exteio a cofigratio of the BASN applicatio. The moel allow to evalate a moitor the trtworthie at compoet, applicatio, a ytem level. I the crret work we are three aitioal apect, applie to trtworthie of ee ata. Firt, we examie i more etail how to erive a opiio (which lea to a level of trt) from oervatio. Seco, we examie the role of certaity i thi, i.e., how to eal with trt-ae eciio whe there i o prior kowlege. We alo propoe a metho to ecay trt over time. Thir, we propoe a way to comie opiio that cotai ch certaity. For thi we e the formalim of jective logic. The ret of the paper i orgaize a follow. Firt we examie the atre of the expecte erroeo ehavior of eor reaig (whether they are malfctioig or exhiitig malicio ehavior) a oerve y the oy h i Sectio
2. Bae o thee typical error we efie a mea to evalate the trt. We explai the operatio of jective logic for thi prpoe a exte it with ew metho to otai opiio from oervatio i Sectio 3. We emotrate the reltig cheme for two propertie of eor reaig for the example of GSR a ECG eor i Sectio 4. Sectio 5 give a overview of the relate work i the literatre. Fially, Sectio 6 give the coclio a ftre work. 2. PRACTICAL ISSUES WITH SENSORS 2. Qality a phyical apect A eor i attache to a aalog-igital-covertor (ADC) of a micro-cotroller, which might tore the reaig to flah or tramit them wirelely. Depeig o the attery level, the pply voltage to the ADC iffer, ch that the eor reaig are affecte. The ADC river affect the timig characteritic with repect to jitter a amplig perio accracy, ch that oie filter (like a 5 Hz filter for ECG) o ot work properly. For high amplig rate, the microcotroller might e ifficiet to proce, tore or tramit the collecte reaig. The atea cofigratio affect the commicatio rage that relt i coectivity prolem. The Galvaic Ski Repoe (GSR) i a mearemet of electrical coctace of the ki that varie with moitre level [4]. The weat gla, repoile for moitre level of the ki, are activate whe a iivial experiece emotioal aroal. Real-life mearemet of GSR poe a mer of challege relate to artifact removal a evalatio of trtworthie of the igal. Artifact i cae of GSR are recorig which o ot tem from emotio-ice chage i weat gla activity. Thee may relt from the recorig procere or from phyiological repoe i ytem other tha the electroermal oe []. Figre epict ome of the mot commo recorig artifact experiece i real-life ettig. The pictre how a example of a GSR ata mearemet from the DTI-2 writa evice [3] with a Hz amplig freqecy. Figre : Example of artifact of GSR ata mearemet: (a) lo of ki cotact; () motio artifact; (c) repoitioig of the electroe. Recorig artifact may tem from propertie a eitivity of the evice that etermie the miimal a maximal vale of the GSR level that ca e relialy meare. Recorig that are too low may relt from lack of coectivity with the ki (ee (a) i Figre ); too high vale ca e a relt of a expecte repoe to a trog timl that excee the mearemet limit. Other type of artifact ca relt from, o calle, caliratio mark. Thee ofte appear a jmp i the recorig, followe y a expoetial aaptatio to the ew aelie. Similar jmp ca alo relt from repoitioig of the electroe o the ki (ee (c) i Figre ). Phyiologically ae artifact may relt from ki movemet irectly eeath electroe, t alo from mclar activity eve qite far from the recorig ite itelf (ee () i Figre ). Aitioal artifact may arie from prere a tretchig of the ki a from chage i ki loo flow. The GSR recorig ca alo e affecte y a trog repiratory activity, gro phyical movemet or participat peech activity. We coier the GSR igal aro thee artifact a le trtworthy; we wat to evelop metho to rig thi ito or trt moel. 2.2 Malicio activitie The proper operatio of the oy h a the eor ca e affecte y malicio activitie [6, 9]. A eial-of-ervice attack ca target the operatio of the oy h y eig exceive ata, lock the commicatio chael, or eplete the attery of the eor y eig malicio reqet. To prete certai oy ehavior, a eqece of eor ata ca e replaye y a malicio eor or tool. I a ooff attack, a eor alterate etwee perio of reportig correct a icorrect iformatio, which i ofte harer to etect ice it might e legitimate. A temperatre or light eor that alterate etwee poitio i haow a i irect light relt i a o-off ehavior. 2.3 Aormal reaig For mot eor, there exit a typical ehavior. For a temperatre eor, the reaig hol e fairly tale whe operate i a home or office, while for a ECG eor, a perioic peak hol occr. The ecrie prolem relt i mearemet which are icoitet with the typical ehavior, for example eviatig from temperatre mearemet, t the ehavior might alo e the relt of kow pheomea for which o cotermeare exit. To etect aormal reaig, a geeric pecificatio of the expecte ehavior i reqire, for example, ae o the maximm variace withi a certai iterval (for temperatre), the expecte variace withi a iterval (for ECG) or the expecte freqecie (i the freqecy omai). 3. TRUST EVALUATION SCHEME 3. Prelimiarie We e the efiitio of trt efie aove for evalatig the trtworthie of eor, which are coiere a trtee while the oy h a applicatio rig o the oy h are trtor. Applyig thi efiitio to the relatio etwee the oy h a eor, the ehavior a tatemet aot it refer to the eor operatio, the ee ata, a propertie thereof. The trt i the a meare of how the oy h jge trtworthie of the eor reaig, while the jtificatio of thi pertai to the comptatio of the trt vale, e.g., ae o moitorig, ata coitecy, reptatio or other mea. We coier the approach for trt evalatio ae o jective logic, i which the trt evalatio proce follow certai evalatio tep to il opiio o qality attrite of the trtee. Opiio are the ai of jective logic a they are e to expre elief. Qality attrite may cotai irect oervatio regarig the
R i D i... Oervatio qality attrite Opiio geeratio Parameter (trt profile) opiio Opiio from other (reptatio) Opiio comiatio opiio (trt vale) Weight vector (trt profile, ytem cotext) Deciio makig Trt policy Figre 2: The trt evalatio cheme with two tep of the opiio geeratio a opiio comiatio. techical performace of the eor (e.g., availaility, ata itegrity) or recommeatio from other etitie. Figre 2 how the trt evalatio cheme propoe with two tep of the opiio geeratio a the opiio comiatio. Opiio geeratio: a opiio of each qality attrite i geerate ae o the irect oervatio of the trtee, e.g., the oy h oerve qality attrite of the eor reaig. Sice each eor ca have a ifferet et of qality attrite a eve a qality attrite ca have ifferet propertie (e.g., the oy temperatre hol e tale withi a certai iterval while a ECG hol ot), a particlar geeratio fctio i implemete for each property or attrite. The opiio geeratio fctio ca alo e ctomize throgh a et of parameter, which are pecifie i the trt profile of the trtor. The vale of thee parameter might e ifferet accorig to a give cotext. Opiio comiatio: the opiio of qality attrite are comie accorig to tatemet aot the trtee ehavior or a expectatio of the trtor. Optioally, opiio of other trtee or trtor (reptatio), which are relate to the give cotext, ca e coiere for the trt evalatio. Uig the opiio of other help to evalate the trtworthie more preciely a to etect more complex aormal eor reaig. The comiatio i ae o a weight factor (calle importat rate i [4]) of the attrite; thi i imilar to or trt evalatio metho propoe i [3]. The weight factor i etermie i the trtor trt profile or i ae o the ytem cotext. A otcome of the fial evalatio i a overall opiio or a vector of opiio, which are compare to a threhol for makig a eciio. The threhol a gielie for eciio makig are pecifie i the trt policy. We igore typical ecrity prolem of meage atheticatio a key itritio. There are may approache a oltio i the literatre for ch prolem. 3.2 Trt evalatio ig jective logic Withi opiio we wat to itigih trt, itrt a certaity. The latter captre a lack of kowlege; a high certaity mea that we ee to fall ack to a li trt. Thi i captre preciely withi jective logic. Sjective logic i a logic that operate o jective elief aot the worl. It coit of a elief moel calle opiio a a et of iary operatio for comiig opiio. It i ite for or prpoe for three reao: ) it pport reaoig from the perpective of a ject rather tha aimig at a ojective otcome, eqal for all ject; 2) it ha mechaim to comie ch opiio; a 3) it allow to take certaity ito accot a ito thi compoitio. Formally, a opiio eote y ωx A = (, a) expree the relyig party A elief i the trth of tatemet x. For example, the opiio of the trth of tatemet Seor S i reliale to provie ee ata with qality q ca e iterprete a trt i S withi the cope of the actio or ehavior provie ee ata with qality q. Here,,, a repreet elief, ielief, a certaity repectively, where [, ] a + + =. Ucertaity i cae y the lack of eviece to pport either elief or ielief. The parameter a [, ] i calle the ae rate, a repreet the elief whe o eviece i kow; i aitio it i e for comptig a opiio expectatio vale efie a E(ωx A ) = + a. Sjective logic preet everal operator to comie opiio (ee [] for etail). For example, coe ( ) comie opiio of ifferet etitie o the ame etity; icot ( ) geerate a opiio of a recommeatio or a chai of recommeatio; cojctio ( ) ece a comie opiio o the cojctio of two itict opiio. A importat oe for or work i the opiio geerator, e to geerate opiio from real worl oervatio o a certai tatemet. Let p e the mer poitive oervatio aot the tatemet, the mer of egative oervatio, ε a parameter cotrollig the rate of lo of certaity, which ca e e to te the e of certaity i the moel for the reqiremet of ifferet ceario, the: = p p + + ε ; = p + + ε ; = ε p + + ε Thi efiitio of opiio geerator ha limitatio: it oe ot take time ito accot a whe evalate over a large mer of vale it tailize. We will ic thee limitatio a or propoe geeratio fctio (e.g. areig a limite hitory of vale) i Sectio 3.3. I aitio to the aove operator, the aig operator ( ) i itroce i [4], which i e to aggregate opiio over a mer of ifferet -opiio. Let QA = qa, qa 2,..., qa } e a et of attrite that may iflece a etity opiio o a tatemet. W = w, w 2,..., w } ( i= wi = ) i a weight vector efiig importace rate of attrite i QA, accorig to the etity coieratio. Ω = ω, ω 2,..., ω } i a et of opiio aot the qality attrite. ω i = ( i, i, i) i the opiio o qa i. The ae rate a i = a for a give cotat a. The aggregate opiio ω = (,, ), i which i efie a = () w i i (2) i= Similarly, a are efie. 3.3 Opiio geeratio The opiio that trtor A (e.g., a applicatio) ha o eor i i eote y ωi A = () (a i a give cotat). To implify the otatio, here we e itea of A i (imilar to a ). The correpoig trt vale i efie a the opiio expectatio vale: tvi A = + a; thi trt vale i i the rage [, ]. The opiio aot each qality attrite of i i compte ae o a et of irect oervatio aot ch qality attrite or other eviece. I
the cope of thi paper, we coier two qality attrite: amplig reliaility (R i) a ata itegrity (D i). We will evelop the opiio geeratio fctio (geerator) for thee qality attrite. The correpoig opiio are eote y ω A R i = ( R, R, R) a ω A D i = ( D, D, D), repectively. 3.3. Samplig reliaility The oy h precrie the amplig perio (p i ) at which eor i hol collect ample. De to practical ie metioe i Sectio 2., the actal amplig perio might eviate from the reqete perio. Give a erie of amplig time τi the ata y i), we efie p i = τi τ i for eor i with (timetamp pt o to repreet the actal ample perio for ample. The accracy of ample i the efie a: p i = p i p i (3) Whe the applicatio reqire a relative accracy of α, the mer of ccefl a faile ample are collecte i r i a rf i, repectively, accorig to the followig formla: if p i α p i r i r i + (4) ele rf i rf i + I (4), relative accracy i e to make a eciio a the complete hitory of eor i i repreete y the vale r i a rf i. With mior ajtmet, the eciio ca e ae o aolte accracy, a the vale of r i a rf i ca pate to rece the effect of a log hitory. For example, the followig formla will e e. if p i α p i r i r i + ; rf i γ rf i (5) ele rf i rf i + ; r i δ r i where γ a δ are i [, ]. The γ a δ vale iicate the level of peimim a optimim a etermie how fat ielief or elief are lot whe face with cce or failre, repectively. With flctatig ccee a failre, oth a, δ γ r a rf remai mall a will coverge to repectively, ch that the certaity factor ε play a more importat role. Whe cce a failre are o-oolea vale i the rage [, ] (e.g. ae o the accracy), the aove formla ca e captre with the followig igle formla. ri δ f r i + rf i γ (6) rf i + f where = α p i /p i a f = ;, f [, ]. For (, f) = (, ) a (, f) = (, ), it relt i the previo formla. The opiio ωr A i of the amplig reliaility attrite i compte from the mer of ccefl a faile ample, accorig to the geeratio operator (ee Eqatio ()). R = r i r i + rf i + ε ; R = rf i r i + rf i + ε ; R = ε r i + rf i + ε (7) 3.3.2 Data itegrity A iicate i Sectio 2.3, the ata ample collecte y a eor will ofte follow a certai expecte ehavior. To etermie the qality attrite for ata itegrity, the oy h e the expecte ehavior for the ample recore y eor i, ae o previo ample. Let ata i repreet the ata tream of ample from eor i, where ata i repreet the vale of ata ample, for. For a tale or lowly chagig igal, the expecte vale expect i will e cloe to the average vale of the lat N ample: expect i = k= N atak i N, for N (8) The mer of previo ample (N) e to compte the average i a parameter. Bae o thi expecte vale the ata itegrity of each ample i etermie. Similar to Sectio 3.3., the mer of ccefl a faile ample are collecte i i i a if i, accorig to the followig formla: if expect i ata i < β i i i i + (9) ele if i if i + where β repreet the maximm allowe aolte eviatio from the expecte vale. The opiio ω A D i of the ata itegrity i the compte y coierig i i a if i, ig the geeratio operator. Thi i imilar to the calclatio i Eqatio (7). The aove qality attrite i relevat for eor that are atrally tale (e.g., the oy temperatre), where the parameter N a β are pecific for the type of eor or the applicatio. For eor ata that are atrally flctatig (e.g., ECG) a ifferet qality attrite ee to e e althogh it ca till e e to etect extreme otlier. 3.4 Opiio comiatio Mltiple opiio or qality attrite, traitive opiio, or opiio from other are comie to a overall opiio (or a vector of opiio). The opiio or recommeatio from other, i a imilar cotext, ca e coiere to icreae the accracy of the evalatio proce or to etect complex aormal ehavior of the eor. For example, the correlatio etwee the ata receive from the iteral accelerometer of the oy h (e.g., a mart phoe) a the ata receive from exteral accelerometer eor ca e take ito accot. The comiatio operator metioe i Sectio 3.2 are e i the opiio comiatio proce. For example, we coier the comiatio of two opiio aot the amplig reliaility a ata itegrity attrite, ωr A i a ωd A i repectively. The overall opiio, Ω A i = (,,, a), i etermie accorig to the aig operator (ee Eqatio (2)); for example i efie a follow. = w R R + w D D () where w R a w D repreet the weight for reliaility a itegrity attrite, repectively. Agai, the weight vector ca e pecifie for each et of qality attrite i the trt profile or it ca e aapte ae o the give cotext. To illtrate the opiio comiatio, we ecrie a example of the tre moitorig applicatio (A), which moitor the tre level from the heart-rate ata. The heartrate ata i provie y the ECG eor (). Let ω A = (,.,.) (the ae rate a =.5) e the opiio of A aot the ata of. B i the epilepy moitorig applicatio that etect epileptic eizre ae o the heart rate. Whe B i italle o the ytem, B e the heart-rate ata of. Ame B ha referral trt i A, eote y ωa B = (,.,.3). I certai precoitio [] it ca e cocle from (B trt A) a (A trt ) that (B trt ). B iirect trt i ca the e erive y icotig A trt i with B trt i A. The erive trt i eote y ω B:A = ( B:A, B:A, B:A ). By ig the icot
GSR [S] 5 5 Time [] 2 4 6 8 2 4 6 Figre 3: The GSR ata mearemet from the DTI-2 writa with 2 Hz amplig freqecy. operator ( ), we have ω B:A ωa B ω A = (8,.6, 6). The effect of icotig i to icreae certaity, i.e. to rece the cofiece i the expecte vale. The applicatio B the ecie to evalate the trtworthie of ae o it irect oervatio aot. Ame that B ha the opiio aot, eote y ω B = (,.2, ). The opiio ω A B = ( A B, A B, A B ) i the calle the coe etwee ω A a ω B, eotig the trt that a imagiary applicatio [A, B] wol have i. By ig the coe operator ( ), we have ω A B ω A ω B = (.78,.3,.9). The coe operator rece certaity, i.e. to icreae the cofiece i the expecte vale. 4. EXAMPLES We emotrate the trt evalatio accorig to two qality propertie (amplig reliaility a ata itegrity) of eor reaig for the example of GSR a ECG ata. Thee ata are collecte from the eor y ig the VIT- RUVIUS oy eor platform [2]. 4. A example of GSR ata Trt vale aic limite hitory o oolea cce Time [].2 2 4 6 8 2 4 6 Figre 5: The trt vale of amplig reliaility for GSR ata accorig to the aic, limite hitory, a o-oolea cce fctio. i, if i, if 4 x 4 (a) Baic fctio: Variale (i) a (if) of cce a failre, repectively 2 i if 2 4 6 8 2 4 6 () Baic fctio: Belief, ielief, a certaity vale.5 2 4 6 8 2 4 6 (c) Limite hitory fctio: Variale (i) a (if) of cce a failre, repectively i 5 if 2 4 6 8 2 4 6 () Limite hitory fctio: Belief, ielief, a certaity vale.5 2 4 6 8 2 4 6 Figre 6: The opiio ω D = () aot the GSR ata accorig to the ata itegrity attrite. p,,,,,, 5 (a) p Time [] 2 4 6 8 2 4 6.5 () Baic fctio 2 4 6 8 2 4 6.5 (c) Limite hitory fctio 2 4 6 8 2 4 6.5 () No oolea cce fctio 2 4 6 8 2 4 6 Figre 4: The opiio ω R = () aot the GSR ata accorig to the amplig reliaility attrite. Figre 3 preet a example of the GSR ata mearemet from the DTI-2 writa [3] with 2 Hz amplig freqecy. I the firt evalatio, opiio aot GSR ata accorig to the amplig reliaility attrite are geerate y ig three ifferet fctio: the aic fctio (ee Eqatio (4)), the limite hitory fctio (ee Eqatio (5)), a the o-oolea cce fctio (ee Eqatio (6)). The followig parameter of the geeratio fctio are choe: ε =, α =.5, γ =.5, δ =.5, a the expecte amplig perio p =.5 (ecae the amplig fre- qecy i 2 Hz). The accracy of ample i term of the ample perio ( p) a the opiio ω R = () are compte a how i Figre 4. From the opiio () the correpoig trt vale (tv = + a, where the ae rate a =.5) are compte. Thee trt vale are how i Figre 5. I the eco evalatio, the opiio are geerate accorig to the ata itegrity attrite y ig the aic fctio (ee Eqatio (9)) a the limite hitory fctio (ee Eqatio (5)). The followig parameter are choe: β = 6 a N =. The mer of ccefl a faile ample (i a if) a the opiio ω D = () are compte a how i Figre 6. From the opiio the trt vale are compte for the cae of the aic fctio a the limite hitory fctio. The correpoig trt vale are how i Figre 7. I the aove evalatio for oth qality attrite, the limite hitory a o-oolea cce fctio are etter to aapt to the qality of the eor reaig. Sice the aic fctio e the complete hitory of the eor reaig a the mer of faile ample are omiate y the mer of ccefl ample i ch example, the vale of () are motly cotat. The elief vale i motly eqal to (thogh the elief ecreae lightly whe there are faile ample, t it i ot viile o the figre ecae of the low reoltio). I cae of the limite hitory (or o-oolea cce) fctio, the variale i a if i Figre 6.(c) o ot reflect the cce a failre occrrece. Thee variale are pate a cotrolle y the parameter γ a
Trt vale aic.2 limite hitory Time [] 2 4 6 8 2 4 6 Figre 7: The trt vale of ata itegrity for GSR ata accorig to the aic a limite hitory fctio. ECG 28 26 24 22 2 5 6 7 8 9 2 Time [] Figre 9: The ECG ata mearemet from the Shimmer eor with Hz amplig freqecy. Trt vale.5 (a) Aggregate opiio of "aic" opiio. Time [] 2 4 6 8 2 4 6.5 () Aggregate opiio of "limite hitory" opiio. 2 4 6 8 2 4 6 aic limite hitory (c) Aggregate trt vale for "aic" a "limite hitory" opiio. 2 4 6 8 2 4 6 Figre 8: The overall opiio a trt vale, ae o comiig the opiio o the amplig reliaility a ata itegrity. p,,,, 3 2 (a) p 5 6 7 8 9 2 () Baic fctio Time [].5 5 6 7 8 9 2.5 (c) Limite hitory fctio 5 6 7 8 9 2 Figre : The trt evalatio for ECG ata ig the amplig reliaility attrite. δ to rece the effect of a log hitory. For example, the vale of i ecreae to δ i whe there i a failre (i.e. the trtor forget part of the cce). The vale of δ etermie whether the trtor i optimitic or peimitic. We ote that thee fctio o ot provie a effective evalatio agait all poile malicio activitie or aormal reaig. For example, to etect a replay attack, a algorithm i reqire to etermie the imilarity etwee recet a hitorical eor reaig. Whe the imilarity i expree a a (o-oolea) cce vale, it ca e e a ipt for a opiio geerator. We emotrate the opiio comiatio y ig the aig operator ( ) to aggregate the overall opiio over two opiio (ω R a ω D). The comiatio i ae o the importace rate of the qality attrite. We ame that the GSR ata i e y a tre moitorig applicatio to etect tre level of a er. From thi applicatio perpective the amplig reliaility (qa R) i le importat tha the ata itegrity (qa D), a the correpoig weight vector of thee propertie i W = (w R, w D) = (.35, 5). Figre 8 plot the overall opiio a the correpoig trt vale, where the opiio for amplig reliaility a ata itegrity are comie for the aic a limite hitory fctio. The overall trt vale provie the applicatio with a etter evalatio of the trtworthie of the ata. For example, i Figre 8.(c), the trt vale of the GSR ata i ecreae at time, which i cae y a rece elief i the amplig reliaility. However, the rece trt i ot viile i Figre 7. 4.2 A example of ECG ata Figre 9 plot the ECG ata mearemet from the Shim- mer ECG eor [2] with Hz amplig freqecy. Similar to the aove example of the GSR ata, we evalate the trtworthie of the ECG ata y coierig the amplig reliaility attrite with two opiio geeratio fctio: the aic a the limite hitory fctio. The followig parameter are choe: ε =, α =.5, γ = 5, δ = 5, a the expecte amplig perio p =. (ecae the amplig freqecy i Hz). The p a the correpoig vale of elief, ielief, a certaity () are compte a how i Figre. The aic fctio relt i a imilar ehavior (the vale of () are motly cotat) a i the example of the GSR ata. The elief a ielief vale are aro.7 a.3, repectively, which iicate alteratio etwee ccee a failre. I cae of the limite hitory fctio, thoe alteratio relt i low vale for r a rf, ch that the certaity factor ε ha a more omiat role. A how i Figre.(c), the certaity get a higher vale tha the elief a ielief. The correpoig trt vale compte from the opiio are how i Figre. 5. RELATED WORK I wirele eor etwork (WSN), trt maagemet i pecifically efl to maage the trtworthie of iteractio amog etwork etitie [5, 7]. The athor i [5] preete a ecrity framework with trt maagemet that i e to ecre eor etwork. A itrite trt moel ealig recommeatio-ae trt a trt-ae recommeatio ha ee propoe, to il reaoale trt relatiohip amog etwork etitie. I [8] the athor propoe a trt maagemet moel for evalatig the trt vale of eaco oe that provie the locatio iformatio
Trt vale aic limite hitory Time [].2 5 6 7 8 9 2 Figre : The trt vale of ECG ata accorig to the aic a the limite hitory fctio. for a BASN. The moel i ae o a itrite reptatioae trt moel, i which each eaco oe tore it pat irect iteractio experiece (i.e. the mer of correct a icorrect locatio roacat from each eighorig eaco oe) locally. Each eaco oe evalate the reptatio of it eighor y comiig it oerve pat eviece a the firt-ha eviece. The BASN the e the reptatio vale receive from eaco oe to compte the overall reptatio (trtworthie vale) of each eaco oe a the locatio iformatio from trte eaco oe will e e to etimate the crret locatio of the BASN. I [9] the athor claime that mot of the trt maagemet moel propoe for MANET a WSN are ot itale for BASN ice they o ot coier the iqe operatioal a ecrity reqiremet of BASN. For example, the trt maagemet protocol i [5] compte trt i a flly itrite maer, i which each eor oe moitor the ehavior a maage the trt recor of other oe. Thee mechaim icr high cot o eor oe i term of proceig power, memory, awith, a eergy comptio. The athor alo ecrie their attack-reitat a lightweight trt maagemet cheme ame ReTrt. The trt i etalihe ae o oth the irect oervatio a recommeatio from other etitie accorig to the mer of ccefl a faile iteractio etwee oe i the etwork. The athor however i ot ecrie how to compte thee ccefl a faile mer. I or propoe framework, the opiio cocept of jective logic i e a the ai for trt evalatio, which take ito accot the certaity factor. Sjective logic alo provie everal operator for comiig oth irect a iirect trt vale. 6. CONCLUSION Thi paper ivetigate the atre of oerve eor error i a oy eor etwork. It propoe a framework to evalate the trtworthie of eor reaig from qality attrite (with amplig reliaility a ata itegrity a example) withot torig log time erie. The trt evalatio i performe throgh two tep: the opiio geeratio a the opiio comiatio. The exteio of jective logic i evelope for thi prpoe. The trt evalatio wa emotrate with two example of the GSR a ECG ee ata. It wa how that the opiio geeratio fctio with the limite hitory i etter to aapt to the qality of the eor reaig. I aitio, the fctio parameter give a trtor the aility to elect optimitic or peimitic ehavior accorig to the ytem cotext. The comiatio of opiio help to evalate the trtworthie more preciely a to etect the complex eor error. Crret a ftre work i o exteig the et of qality attrite th pecifyig aormal reaig more preciely for ifferet type of eor. Moreover, cae-etectio a cotrol mechaim will e ivetigate. Feeack loop ig thee mechaim ai to maitai the trtworthie. Ackowlegmet We wol like to thak IOP GeCom a the VITRUVIUS project for the fiacial pport for thi work. 7. REFERENCES [] W. Bocei. Electroermal activity. Spriger, 2. [2] V. Bi, R. Verhoeve, a J. Lkkie. A oy eor platform for cocrret applicatio. I IEEE It. Cof. o Comer Electroic - Berli (ICCE-Berli), page 38 42, 22. [3] V. T. Bi, J. J. Lkkie, a R. Verhoeve. Towar a trt maagemet moel for a cofigrale oy eor platform. I Proc. 6th It. Cof. o Boy Area Network (BoyNet 2), page 23 26, 2. [4] C. W. Darrow. The ratioale for treatig the chage i galvaic ki repoe a a chage i coctace. Pychophyiology, ():3 38, 964. [5] S. Gaeriwal, L. K. Balzao, a M. B. Srivatava. Reptatio-ae framework for high itegrity eor etwork. ACM Traactio o Seor Network (TOSN), 4(3): 37, Je 28. [6] S. Gaeriwal a M. B. Srivatava. Trtworthy eor etwork: Ie a challege. Techical report, UC Lo Agele: Ceter for Emee Network Seig, 24. [7] L. Gomez, A. Lae, a A. Soriotti. Trtworthie aemet of wirele eor ata for ie applicatio. I It. Cof. o Avace Ifo. Networkig a Applicatio, page 355 362, 29. [8] Z. S. Ha Y a C. Leg. Towar trt-aware health moitorig oy area eor etwork. It. Joral of Iformatio Techology, 6, 2. [9] D. He, C. Che, S. Cha, J. B, a A. V. Vailako. Retrt: Attack-reitat a lightweight trt maagemet for meical eor etwork. IEEE Tra. o Ifo. Tech. i Biomeicie, 6(4):623 632, 22. [] A. Jøag. A logic for certai proailitie. Iteratioal Joral of Ucertaity, Fzzie a Kowlege-Bae Sytem, 9(3):279 3, Je 2. [] A. Jøag, R. Haywar, a S. Pope. Trt etwork aalyi with jective logic. I Proc. 29th Atralaia Compter Sciece Coferece, ACSC 6, page 85 94, Atralia, 26. [2] Shimmer reearch. http://himmer-reearch.com/. [3] J. Weterik, M. Owerkerk, G.-J. e Vrie, S. De Waele, J. va e Eereeem, a M. va Bove. Emotio mearemet platform for aily life itatio. I 3r It. Cof. o Affective Comptig a Itelliget Iteractio, page 6, 29. [4] Z. Ya a C. Prehofer. Atoomic trt maagemet for a compoet-ae oftware ytem. IEEE Tra. o Depeale a Secre Comp., 8(6):8 823, 2. [5] Z. Yao, D. Kim, I. Lee, K. Kim, a J. Jag. A ecrity framework with trt maagemet for eor etwork. I t It. Cof. SecreComm, page 9 98, ept. 25.