waytotestclaimsorideasaboutagroupor

waytotestclaimsorideasaboutagrouporpopulation.
GivenobservationsoftransmissiondataratesubsetsΩ,we
participationΩinton+1subsets{R1,R2,...,Rn,Rc},
correspondingtopossibledataratesetsduetovaryingchannel
conditions,whilethe{Rc}representsdataratesetcausedby
covertcommunications.Thedetectionproblemisformulated
asfollows:
Hp=δ(D)=
⎧
⎪
⎪
⎪
⎨
⎪
⎪
⎪
⎩
H1normal
H2normal
...
Hnnormal
Hccovertchannel
(1)
Here,Hi,i=1,2,...ncorrespondtooneofNULLhy-
pothesis;thealternatehypothesisis{Hc}.Hpisthepredicted
hypothesis.Thechallengehereistofinddecisionruleso
thatthepredictionerrorisminimizedbetweenthepredicted
HpandHi,i=1,2,...nincludingHc.Discriticalvalue.
Wegrouptheoriginalobservationofdataratesetsintotwo
biggroups,arbitrarilylabeledAandB,whicharedifferent
withrespecttosomenumericalcharacteristicsX.InEq.(1),
Hi,i=1,2,...nisconsideredasgroupAandHcisgroup
B.Byanalyzingstatisticsfeatureoforiginaldataratesets,a
featuringvector,consideredasXisformed,andhypotheses
areformedaswell.Inthefollowingsectionwewilldiscuss
howtomapdifferentdataratesets{R1,R2,...,Rn,Rc}to
thecorrespondingHi,Hc,i=1,2,...n;thecriteriaDfor
makingadecision,andthetestprocedure.
C.Theproposeddetectionmodel
802.11a/b/g/nprotocolshavedifferentavailabledatarates
asshowninTableI.Byvaryingmodulationtypes,eachproto-
colallowsadevicechangeitsdataratetoaccommodateunsta-
blewirelessconditions.Rateadaptionisthedeterminationof
theoptimaldatatransmissionratemostappropriateforcurrent
wirelesschannelconditions.Itconsistsofassessingchannel
conditionsandaccordinglyadjustingthetransmissionrate.
Adetailedsurveyonrateadaptionalgorithmscanbefound
in[11].Itisverycommonforadevicetousedifferentdata
ratestotransmitdatainaWLAN,thoughthiscanbeutilized
byanattackertopasssecretmessage.However,justbylooking
atdifferentdataratesusedinaWLANwillnothelpusidentify
anyabnormalcase.Inthispaper,theprobabilitydistribution
oftransmissiondataratesRisproposedtocharacterizerate
diversityinaWLAN.Thischaracteristicsiscalculatedbased
onmeasurementsofdataratesoveracertainperiodoftime.
Thedistributionofdataratesrevealeitherunstablewireless
channelconditionsorapossiblecoverttimingchannel.To
characterizethedynamicfeatureofrateswitchinginaWLAN,
Afeaturevectorisproposedasfollows:
VF=[PR(1),PR(2),...PR(k)](2)
VFisakdimensionalvector,andeachdimensionrep-
resentsaprobabilitydistributionofdatarateRkassuming
therearekdifferentdataratesinaWLAN.Foraspecific
WLAN,notalldataratesareavailable.Inthatcase,missing
datarateisrepresentedby0.NowthestatedhypothesisHiin
Eq.(1)couldbeformedbyVF,representingapossibledata
ratedistributionscenario.EachhypothesisinEq.(1)should
beavectorandboldfaceletterswillbeusedforhypotheses
intheremainderofthispaper.Hicouldbetrainedthrough
realtrafficdata.Hcrepresentsratedistributionpatterncaused
bycovertchannels.Asecretmessageusuallyisnotencrypted
whenitiscarriedbycovertchannelandtheASCIIisencoded
directlybydifferentdatarates.ThehypothesisHcisformed
byletterfrequenciesorwordfrequencies.Letterfrequencies,
howevertendtovarybothbywriterorbysubject.Noexact
letterfrequencydistributionunderliesagivenlanguage,since
allwriterswriteslightlydifferently.Accurateaverageletter
frequenciescanonlybegleanedbyanalyzingalargeamountof
representativetext.Withtheavailabilityofmoderncomputing
andcollectionsoflargetextcorpora,suchcalculationsare
easilymade.TheletterfrequencyfromPavelMika’swebsite,
whichcitesRobertLewand’sCytologicalMathematics,is
adoptedinthispaper.Fig.2showsrelativeletterfrequency.
Arandomtextfileiscreatedwithletterdistributions
followingtheletterfrequencyshowninFig.2.Theselet-
tersarerepresentedbytheirASCIIcodes,whichgiveus
asequencea0sand1s.Dependingonavailabledatarates
inaWLAN,anattackermaypick4differentdatarates
suchasdataratesin802.11b[1,2,5.5,11]Mbpstocodea
secretmessage.Thereare4!=24possibleencodingschemes
toencode00,01,10,11.For8datarates,thenumberof
encodingschemescouldreach8!=40320.Forexample,
datarates1,2,5.5,11Mbpscouldencode00,01,10,11,re-
spectively,whichisoneofmanypossibleencodingschemes.
Thedistributionsofdataratescanthenbecalculatedbasedon
arandomlygeneratedtextmessage.ThusHcrepresentsrate
distributioncausedbycoverttimingchannelsusingdifferent
datarates.TogetaccurateestimationofHc,alargetext
ortheaverageofseveraltextscouldbeusedtoformHc.
FortheHypothesisHi,i=1,2,...n,experimentalmethodis
usedtocharacterizedifferentWLANscenarios.Nowweneed
toproposeacriteriauponwhichwedecidetheclaimbeing
testedbelongstowhichhypothesis.Ifthetestedclaimbelongs
toanyoneofHi,i=1,2,..,nhypothesis,itisanormal
WLANratedistributionscenarioanddoesnotcontainsecret
information;ifitfallsinHc,secretmessageisembedded
intransmissiondatarates.Thus,coverttimingchannelis
detected.Aruntimemonitorcollectssampledataratesfrom
aWLAN;theratedistributionisthencalculatedtoformHp,
whichisakdimensionalvector.Nowwecomparewhatwe
observeinHptowhatweexpectHi,i=1,2...n,Hcinterms
ofsimilarity.ThecriteriaissetbasedonwhichvectoramongHi,i=1,2...n,HcisthemostsimilaronetotheHp.TheEuclideandistanceisproposedtomeasuresimilarity:D(k)=????m

結果 (中文) 3:

waytotestclaimsorideasaboutagrouporpopulation。givenobservationsoftransmissiondataratesubsetsΩ，我们参与该Ω+ 1subsets { R1，R2，…，Rn、RC }，correspondingtopossibledataratesetsduetovaryingchannel条件，而钢筋混凝土representsdataratesetcausedby { }covertcommunications。thedetectionproblemisformulated如下：惠普=δ（D）=⎧⎪⎪⎪⎨⎪⎪⎪⎩h1normalh2normal…hnnormalhccovertchannel（1）在这里，嗨，i = 1,2，…ncorrespondtooneofnullhy—假设；thealternatehypothesisis { } hpisthepredicted HC。thechallengehereistofinddecisionruleso假说。thatthepredictionerrorisminimizedbetweenthepredictedHpandHi，i = 1,2，…nincludinghc.discriticalvalue。wegrouptheoriginalobservationofdataratesetsintotwo大，arbitrarilylabeledaandb，whicharedifferentwithrespecttosomenumericalcharacteristicsx。不等式（1），嗨，我nisconsideredasgroupaandhcisgroup = 1,2，…B. Byanalyzingstatisticsfeatureoforiginaldataratesets，一个featuringvector，consideredasxisformed，与研究假设areformedaswell。inthefollowingsectionwewilldiscusshowtomapdifferentdataratesets { R1，R2，…，Rn、RC }来thecorrespondinghi，HC，i = 1,2，…n；thecriteriadformakingadecision，andthetestprocedure。C. theproposeddetectionmodel802.11a/b/g/nprotocolshavedifferentavailabledataratesByvaryingmodulationtypes，eachproto asshownintablei—colallowsadevicechangeitsdataratetoaccommodateunsta—blewirelessconditions。rateadaptionisthedeterminationoftheoptimaldatatransmissionratemostappropriateforcurrentwirelesschannelconditions。itconsistsofassessingchannelconditionsandaccordinglyadjustingthetransmissionrate。adetailedsurveyonrateadaptionalgorithmscanbefound在[ 11 ]。itisverycommonforadevicetousedifferentdataratestotransmitdatainawlan，thoughthiscanbeutilizedbyanattackertopasssecretmessage。然而，justbylookingatdifferentdataratesusedinawlanwillnothelpusidentifyanyabnormalcase。本文theprobabilitydistributionoftransmissiondataratesrisproposedtocharacterizeratediversityinawlan。thischaracteristicsiscalculatedbasedonmeasurementsofdataratesoveracertainperiodoftime。thedistributionofdataratesrevealeitherunstablewirelesschannelconditionsorapossiblecoverttimingchannel，characterizethedynamicfeatureofrateswitchinginawlan，Afeaturevectorisproposedasfollows：VF = [ PR（1），PR（2），…PR（K）]（2）vfisakdimensionalvector，andeachdimensionrep—resentsaprobabilitydistributionofdataraterkassumingtherearekdifferentdataratesinawlan。foraspecific无线局域网，notalldataratesareavailable Inthatcase失踪。datarateisrepresentedby0.nowthestatedhypothesishiinEq.（1）couldbeformedbyvf，representingapossibledataratedistributionscenario。eachhypothesisineq。（1）应beavectorandboldfaceletterswillbeusedforhypothesesintheremainderofthispaper。hicouldbetrainedthroughrealtrafficdata。hcrepresentsratedistributionpatterncausedbycovertchannels。asecretmessageusuallyisnotencryptedwhenitiscarriedbycovertchannelandtheasciiisencodeddirectlybydifferentdatarates.thehypothesishcisformedbyletterfrequenciesorwordfrequencies。letterfrequencies，l howevertendtovarybothbywriterorbysubject。letterfrequencydistributionunderliesagivenlanguage，自allwriterswriteslightlydifferently。accurateaverageletterfrequenciescanonlybegleanedbyanalyzingalargeamountofrepresentativetext。withtheavailabilityofmoderncomputingandcollectionsoflargetextcorpora，suchcalculationsare很容易地就交上了。theletterfrequencyfrompavelmika"swebsite，whichcitesrobertlewand"scytologicalmathematics，是adoptedinthispaper.fig.2showsrelativeletterfrequency。arandomtextfileiscreatedwithletterdistributionsfollowingtheletterfrequencyshowninfig.2.theselet—tersarerepresentedbytheirasciicodes，whichgiveusasequencea0sand1s。dependingonavailabledataratesinawlan，anattackermaypick4differentdataratessuchasdataratesin802.11b [ 1,2,5.5,11 ] mbpstocodeasecretmessage.thereare4！= 24possibleencodingschemestoencode00,01,10,11 for8datarates，数量。encodingschemescouldreach8！= 40320.forexample，datarates1,2,5.5,11mbpscouldencode00,01,10,11，重新—性，whichisoneofmanypossibleencodingschemes。thedistributionsofdataratescanthenbecalculatedbasedonarandomlygeneratedtextmessage.thushcrepresentsratedistributioncausedbycoverttimingchannelsusingdifferentTogetaccurateestimationofHc，alargetext数据速率。ortheaverageofseveraltextscouldbeusedtoformhc。FortheHypothesisHi，i = 1,2，…N，experimentalmethodisusedtocharacterizedifferentwlanscenarios。nowweneedtoproposeacriteriauponwhichwedecidetheclaimbeingtestedbelongstowhichhypothesis。ifthetestedclaimbelongstoanyoneofhi，i = 1,2，..，nhypothesis，itisanormalwlanratedistributionscenarioanddoesnotcontainsecret信息；ifitfallsinhc，secretmessageisembeddedintransmissiondatarates。因此，coverttimingchannelisaruntimemonitorcollectssampledataratesfrom检测。theratedistributionisthencalculatedtoformhp无线，whichisakdimensionalvector。nowwecomparewhatweobserveinhptowhatweexpecthi，i = 1,2…n，hcinterms相似性。ThecriteriaissetbasedonwhichvectoramongHi，i = 1,2…n，HcisthemostsimilaronetotheHp.TheEuclideandistanceisproposedtomeasuresi