SecurityAnaly.cs:Real-.mecapabili.esandthepathto‘MinorityReport’

BenjaminTan(A0120523Y)StephenDay(A0120684J)HuSong(A0027392E)YongHow(A0120654N)

MT5009:AnalysingHigh-TechFinalProjectPresenta;on

TheStoryboard

SeSngtheScene(Introduc.on)• SubjectArea&Objec.vesforthePresenta.on• ProblemDefini.on–Real-.meRequirements

Wri.ngtheScript(TheResearchandMarket)• BigDataandAnaly.csMarkets• OverviewofthesecuritymarketforReal-.meSecurityAnaly.cs

OnSetwiththeActors(Today’sSolu.ons)• BigData&CloudThreatIntelligence• Sensorsandstreamingdataforsecurityapplica.ons

PreparingfortheSequel(FutureProjec.onsforTechnology)• AIandmoreintelligentsystemsthatdeliverprescrip.veac.ons• Sensorprolifera.onandnewendpoints

It’saWrap(Summary&Conclusions)

Introduc.on:Technologies&EntrepreneurialOpportuni.es

•  DataStreamingArchitectures•  Analy.csAlgorithms&Insights•  CloudInfrastructureandHyperscale

Compu.ng•  Sensors(Video,Voice,Image,

Biological,Chemical,GPS)•  Plaeorms(IoT,Drones,HetNets)•  SecurityOpera.onsCentres•  Prescrip.veAlertResponseteams

MarketFocus:Int’l&Domes.cSecurity&IntelligenceAgencies

Situa.on:22ndMarch2016BrusselsAfacks

7:58am,Airportbomb

~9am,MetroSta.onbomb

3:00pm,surveillancevideo

discoveryofanewexplosivedevicecontainingnails

6:00pm

•  Mismatched&SiloedIntelligencecollec.on

•  Retrospec.veIntelligenceanalysis

•  Dataisbeingcollectedbutstored•  Notreal-.meandnoinsight

genera.on

ProblemStatement:PriorIntelligenceFailed

Policeraidsaweekagohitoneoftheterroristsapartmentandkilledagunmanandcapturedonealive(SalahAbdeslam)linkedtoParisafack

“TheBelgiansecurityservices,aswellasotherWesternintelligenceagencies,hadadvanceandpreciseintelligencewarningsregardingtheterroristafacksinBelgiumonTuesday”-Haaretz

“U.S.intelligenceofficialssaytheyweren'tsurprisedthatBrusselswasafacked,becauseofconcernsaboutterrorthreats,par.cularlyaperrecentraidsandthearrestofParisafackssuspectSalahAbdeslamlastweekinBrussels”-CNN

“ThesourcessaidthatKhalidElBakraouiandBrahimElBakraouiwerebothonU.S.governmentcounterterrorismwatchlists”-Reuters

ElBakraoui,wasalreadycaughtinJune2015nearTurkey'sborderwithSyriaandaskedtobedeportedtotheNetherlands.

Solu.onDefini.on:Real-.mePrescrip.veIntelligenceü  Real-.meDatacollec.onü  Real-.meDataprocessingü  Real-.meThreatAnaly.csü  Real-.mepredic.ve&prescrip.vealerts

Now Future

ANALYTICSMARKET

Analy.csMarket:UsageProfiles

Source:Forrester’sGlobalBusinessTechnographicsDataAnaly;csSurvey,2015and2014

Analy.csMarket:Patent&VentureFundingSpeaks

Source:CiscoTechnologyRadar(2014)

Analy.csMarket:SecurityAnaly.csToday

In2012,businessintelligence,corporateperformancemanagement,andanaly.csapplica.onsandperformancemanagementsopwarewasworthUS$13.1billionworldwideSource:CiscoTechnologyRadar

Note:Thisdoesnotincludethe‘Black’spendingbyNSA,GCHQandotherIntelligenceagencies.

Analy.csMarket:Analy.csMegaTrend

Source:IBM,Cisco,Frost&SullivanAnalysis

90%ofthedataintheworldhasbeencapturedinthelasttwoyearalone!

Analy.csMarket:SecurityAnaly.csinFigures

Youcannotputa$$$onpeople’slivesandsafety

ANALYTICSTECHNOLOGY&ARCHITECTURES

TechnologyReadiness:BigDataAnaly.csintheHypeCurve

DataArchitectures:TheStateoftheTechnologyToday

Source:ScaleOutSoLwareInc

DataArchitectures:RealTimeandBatchProcessing

Source:CLOUDSECURITYALLIANCE–Bigdataanaly;csforsecurityintelligence

Real-.me(Spark&Beyond)X

DataArchitectures:BigDataEcosystem(MassiveOpenSourced)

•  OriginatedfromGoogle’sOpenSourceHadoopSystem•  EvolvedintoApacheOpenSourceEcosystem•  FuturewillbeHyperscale,HyperstreamingDatacoupledwithAI

DataArchitectures:ComplexDataFlowManagement

•  ApacheNifi(FormerlyNiagarafiles)isaDataFlowOrchestra.onTool•  DonatedbyNSAinDec2014•  Usedforprocessing‘ForeignIntelligence’•  Lowlatency,highthroughput•  User-friendlyGUIforsecurityoperatorsandanalysts•  SupportsIoTinterac.vecommunica.on

DataArchitectures:Real.me(Streaming)DataProcessing

•  Real.mebigdatastreamingframework–  Harnessesin-memoryandclustercompu.ng–  Itera.veinnaturesolendsitselfwelltoMachineLearning

•  Sparkperform3Xfasterusing10xfewermachinethanMapReduce(Databricksbenchmark).

StreamProcessinginmicrobatches

TrueStreamingprocessing

?2017 2020

DataArchitectures:StreamingAnaly.cs

•  StreamprocessingenableMachineLearningalgorithmsthat‘learn’incrementallyoverthedatainstreamingtoenablereal-.mepredic.on/scoringofthreats.

•  Algorithms&UseCases(MachineandDeepLearning)Ø  SupportVectorMachine-AnomalyDetec.onØ  BaysianNetworks–AnomalyDetec.onØ NeuralNetworks–Imagerecogni.onandpredic.vecapabili.esØ  ClusteringTechniques–Similarbehaviours/characteris.csØ NetworkGraphs–Affilia.onsandinforma.onflowsØ  TextAnaly.cs–Keywords,sen.ments,categorisa.onofinforma.onØ  TimeSeries–Predic.vebehaviourwithoutseasonalitytrends

Analy.csAlgorithms&Insights:AdvancesinDeepLearning,Analy.cs&AI

Source:www.predic;veanaly;cstoday.com

Prescrip;veAnaly;cs

FuzzyTransi;on

ü  Con.nuousqueriesonunstructuredandstructurestreamingdata

ü  Generateac.onableeventtoallowimmediatecontrolandadvanceddecision

Analy.csAlgorithms&Insights:Real-.meVisualisa.on&DynamicNetworkGraphs

•  Machinelearningandalgorithmsgenerateintui.vevisualisa.ons

•  Generatehotspotsoftraffic,connec.ongraphsandiden.fieskeythreatnodes

CloudInfrastructure&HyperscaleDataCentresHyperscalecompu.ngandSopwareDefinedNetworkingarethe2maindrivingforcesofCloudInfrastructuregrowth

•  IntelBroadwell22corechip•  LenovoHXhyperscale

servers•  Nutanixsopware

SENSORS&DATACOLLECTION

Sensors&DataCollec.on:SensorNumbers&Cost

Trajectoryisheadinginexactlytheright

direc.onfordisrup.vedeployment

Sensors&DataCollec.on:SensorTechnologies&Types

Sensors&DataCollec.on:VideoSensors+Analy.csNow

•  Videoanaly;csareusingthelatestmodellingtodothefollowing:–  FacialSurveillancetodetectknownsuspects–  DwellandLoiteringTimetohighlightpoten.alerrantbehaviour–  PeopleCountsforthroughputflowandcrowdcontrol–  Demographicsforsurveillanceandprofiling

•  BeTermechanical,op;calandelectronics–  Sharperimageswithgrowinglevelofdetail–  Sosmallthatitshardlyno.ceablewiththenakedeye

•  Integra;onofprivateandpublicmonitoringcameras–  Mul.-persontrackingcapabili.esanddynamicimageryanalysis–  3-Dimensionaltechniquestoiden.fyanindividualwithinamobon

themove–  Videosarecollatedontothecloud–  Nomoreindividualbutcollec.veanalysisofvideossimultaneously

•  AppearanceofevensmallerCMOSimagesensorarrays•  Developmentofdistributedwirelesssensornetworks•  DistributedprocessingofimagedataonaGPU•  Analy.csprocessingdoneonthecameraandonlyrelevant

metadatasendbacktotheuser•  Datafusionfrommul.pleimagesourcessentbacktouserfor

processingwithloca.onandtargetbasedmetadata•  Efficiencyincommunica.onnetworkresources

Sensors&DataCollec.on:VideoSensors+Analy.csinDevelopment

Sensors&DataCollec.on:VideoSensorCost&Performance

Sensors&DataCollec.on:Predic.veBehavioralAnaly.cs

•  Baseline:‘Normal’ac.vi.esobservedattheloca.on•  Anomalies:Devia.onsfromthatnormthatrequirefurtherexplora.on

–  Smartenoughtorecognizeworkdaysandotherrecurringpaferns–  Onlypickupsta.s.calrari.esofindividualswithinthearea–  Measureagainstthehistoricalpafernsofindividualsinthatarea

•  Predic;ve:forecastbehaviourofindividualsbasedontheanalysisofmul.plepafernsofsimilarpastevents

•  MetaDataAnnota;on:fully-automatedseman.c-taggingofsurveillancevideorecordingsbasedonmul.plehumanpresencedetec.onandabnormalac.vi.esrecogni.on

•  EventDetec;on:automa.csor.ngofvideoandcommunica.onstodetectaggressivehumanbehaviour,publicareasinpar.cular

Sensors&DataCollec.on:IOTandotherSensorProlifera.on

Environmental• Weather-HeatCanpredictcrimewaves• Wind–CyclonesandTyphoonscreatesecurityhazzards

Biological&Biometric•  Agentdetec.on–SarinandAnthraxusedbyterrorists•  Fingerprints,Odour,Perspira.on&DNA–Personaliden.fica.on

Chemical&Nuclear• Agentdetec.on–DirtyBombs,MustardandChlorineusedbyterrorists• Pollu.on-EnvironmentalandHealthhazards

Voice&Text•  NLP&Seman.cSearch(Mul.-lingual)•  Speechtotext,real-.metransla.on,voicestressanalysis

MobileSensors•  Things(Drones,Cars,Buses,Planes…..)•  People(Smartphones&WearableTech)

SensorInfrastructure:HetNets(HeterogeneousNetworks)

•  Mul.plewirelessfrequencies•  Seamlessconnec.vityevery.meandeverywhere,indoorsandoutdoors•  Benefits

–  Efficient,PervasiveandHighQualityCoverageinUrbanareas–  EnhancedResiliencyandSecurityforcri.calinfrastructureandsensors–  Op.misedforsensorcommunica.onrequirements(Lowpower,highspeed,sleep

modes)

ChallengesandEnablers:Deliveringthe‘MinorityReport’

Challenges•  Developmentofnewframeworksfora

GlobalSecurityArchitecture•  PeopleSkills&Trust•  Communica.onvulnerabili.es•  PersonalPrivacyRequirements•  Cost$$$$$

Enablers•  Cheap,fastmemory,storage,compute

andcloudinfrastructure•  Network&Sensorubiquity•  Streamingtherightdata&metadata•  Cheapercloudbasedanaly.cs•  Biometricsecuritymeasures

FUTUREPROJECTIONS

FutureProjec.ons:Prescrip.veAnaly.cs

•  Ar.ficialIntelligence•  Cogni.veLearningNetworks•  AugmentedRealityAides•  Predic.veGameTheory•  Simula.on&ThreatModelling•  DecisionRecommenda.on

Vs.

‘Pre-cogni.ves’ IBMWatson

Conclusion&Summary

Requirement–Real.mePrescrip.veAlerts&ThreatIntelligence

Market–Spend,startupsandinvestmentisstrong

Technology–BigDataCloud,HetNet,Sensors,Analy.csetc

CostRed’n–Chips,Servers,Storage,Comms,Sensors,s/w

PerformanceDrivers-Chips,servers,storage,Comms,sensors,s/w

Sopwarealgorithms–Streaming,AR,AI,Text&Language,Recogni.on6

1

2

3

4

5

ThankYou