Transcription
Make Way for The Internet of Things!SESSION ID: TECH-R02Benjamin JunVP and Chief Technology OfficerCryptography Research, Inc. a Rambus Company
The Internet of ThingsUniquely identifiable objects and their virtualrepresentation in an Internet-like structure.– WikipediaThe physical world is becoming a type of informationsystem [with] sensors and actuators embedded inphysical objects When objects can both sense theenvironment and communicate, they become tools forunderstanding complexity and responding to it.– McKinsey & Company#RSAC2
Brought to you by Compute revolution (80’s)Sensor revolution (90’s)Wireless revolution (00’s)Human Internet (http v1.0 1996)#RSACIntel 4004; ADXL3353
the promisebut Smartgrid smart home energy efficiencyCritical utility DoS? Data collected from many sources andInvasion of privacy?analyzed to gain new insights Physical world modified for the userBurn down house? Real-time marketplace adaptation to dataManipulate markets?#RSAC4
2017: M2M connections x3, traffic x202020: 30-50 billion connected IoT#RSAC5Source: ABI / Cisco / Mocana
Anticipating change50 year-old stove50 year-old computerNCSU Libraries’ Digital Collections: Rare and Unique Materials;http://doowicky.blogspot.com/6#RSAC
Security challenges Connectivity scale huge attack span Ownership is different (BYOD on steroids) Device lifecycle is different Who owns data and device credentials?“Zero-step” activation and M2M transactionsModularity enables future applications Source code forMorris Internet WormBut we don’t know what the threat models are!#RSAC7Photo credit: Shannon B, GoBostonCard.com
We cannot use PC / IT security as the model PC’s continuously updated PC’s have high security investment IoT nodes have long service life!Embedded systems have little or nosecurity support, starting with the SoC BSPIncremental value of PC node IoT nodeNo party willing to spend PC’s have good UI, high user mindshare#RSAC8Gizmodo 8/14/2013; Nitesh Dhanjani
Data at RestData in TransitTime and PlaceEndpoint Security
Phase 1: The database of things!Machine collectedInternet interpretedHuman / machine renderedTileWazewunderground10USGS netquakes#RSAC
What can utility data tell us?Multimedia Content Identification ThroughSmart Meter Power Usage Profiles (Greveler, Justus, Loehr)11#RSAC
Data fusion / Big dataBy 2025 Internet nodes may reside in everyday things Streamlining—or revolutionizing—supply chains andlogistics could slash costs, increase efficiencies, andreduce dependence on human labor. Ability to fusesensor data from many distributed objects could detercrime and asymmetric warfare. Ubiquitous positioningtechnology could locate missing and stolen goods. Massively parallel sensor fusion may underminesocial cohesion if it proves to be fundamentallyincompatible with Fourth-Amendmentguarantees against unreasonable search.Global Trends 2025, US National Intelligence CouncilSupply chainResource efficiencyEmergency servicesCustomizationIntelligence gatheringPrivacyOverstepping controlPredictive “creepiness”PaparazziData poisoning#RSAC12
“Classic” database security issuesConcernExampleData ownershipEuropean Communication COM (2012) 9Data privacyhome occupancy data PCI PII?Data theft“Home addresses recovery PIN for all users of electroniclock model SU-214”Data extractionFacial recognition city camerasExample:PCI regulated dataPCI DSS Requirements and Security Assessment Procedures, v1.213#RSAC
We can’t (yet) manage partial data exposureGraph theory and Facebook (2009)Challenge: partial “peeks” may leak too much#RSAC14
Who holds the data?Centralized data provider Distributed data (in research)Small # of service providers “Security by policy” for dataownership / control / usage spent on quality, security Data monetization a focus Data owners maintain “control” ofcloud based dataFine grained control enforced bycrypto, security protocolsWho will pay for this?802.11s - mesh networkingInternet routingRoot CA#RSAC15
The path to database securityNEAR Data clearinghouses will emerge European privacy requirements business need to aggregate“Security by SLA”Devices encrypt data with user keys “Dropbox” for crypto-partitioned IoT dataRequires device credential & key managementCraig GentryHomomorphic Encryption, MIT Technology ReviewFAR Don’t hold your breath (yet) Encrypted data searchHomomorphic encryption#RSAC16
Data at RestData in TransitTime and PlaceEndpoint Security
Connectivity: My garage doorArduino Ethernet sensors relayEthernet8 bit uP32KB flash2KB RAM#RSAC18
Device as server: Device talks to everybodyUser authentication is abig problemNeed infrastructure-classstability with nomaintenanceRequires hardenedweb serverMore than a web server:Data sharing requiresM2M connectionsDependent on othernetwork resources(time, DNS, DHCP, )#RSAC19
Device-to-cloud: Device talks to one service Theory: Plug-in, VPN directly to cloud server Practice: Infrastructure challenges Nest LabsNot easy to build device that can connect for 20 yearsComplexities (WiFi passwords, TLS certificate expiration, DNS, IPv6, )Everything via VPN?What about hacked/spoofed device (PlayStation Network)Still may require direct device-to-device connections#RSAC20
Device gateway: Device – Gateway – Cloud Gateway to aggregate sensors, actuators Security model firewall to keep bad guys out Bridge Internet to low-power sensor-friendly protocols“NAT for sensors”SCADA gatewaysWhat’s our track record of “inside good” security models?Complexity grows (accumulation of legacy protocols, devices) which brings security bugs(Insecure) example: Vehicle TPMS gatewaysEtherPoll; Abby systems; http://www.cse.sc.edu/ wyxu/papers/TPMSUsenix.pdf21#RSAC
Connection security requires identity management We want global addressability and globalaccessibility What’s in a name? with appropriate controls!Device credentialsIdentify-specific keys, certsWho gets to name it? When? Domain owner, certification authority, issuer, device manager, #RSAC22
IoT protocols to watch Internet of things projects MQ Telemetry Transport (MQTT) Eclipse M2M Industry working groupSecurity to follow OAuth 2.0IM messaging security (Off-theRecord, )Eclipse M2M Industry working group#RSAC23
Data at RestData in TransitTime and PlaceEndpoint Security
Time and place Value proposition: Compute-domain awareness ofphysical thingsAssociations are important The milk bottle in my refrigerator expired Football game will add 8000 more cars to highway at 3:35pm I am standing next to my assigned car-sharing vehicle#RSAC25
Binding to time and place Local services Pay-per-impression Region pricing Relative proximity User mobility User identity 85 for time-based energy demand-response#RSAC26
TimeDisney iOS “Frozen” game timeoutNetflix local clock trackingCommon time sources Local battery User NTP server (pool.ntp.org) Broadcast: GPS, GSM, NIST WWV/WWVH EEPROM (advance only)#RSAChttp://mike-thomson.com/blog/?p 21027
Place (GPS)CJ6 GPS Jammerjammerstore.comOn the Requirements forSuccessful GPS Spoofing AttacksTippenhauer, Pöpper, Rasmussen, CapkunCaptured RQ-170 SentinelChristian Science Monitor, 12/15/2011#RSAC28
Three things the world needs Protocols to selectively prove you were somewhere at some time Secure means to federate devices in close proximity Need: User authentication, proximity based automatic enrollmentConcern: Can I turn off the tracking bug?Two strikes: today’s tech is not private and very spoofable auto pair milk and fridge!Trusted time and location From application OR from serverSamsung SIIIadvertisement#RSAC29
Crypto to the rescue? Trusted Computing Group made some inroads inattestation privacy TPM v1.1 pseudonymous machine credentials (requires TTP)TPM v1.2 direct anonymous attestationNot much infrastructure exists for pseudonymous modes, stillproblematic in real world use scenarios (revocation)#RSAC30
Coming soonTime & place attestation without user / OS / application trust Approach 1: Chipsets w/ built-in environment attestation resource Independent core on CPU maintains GPS time historyHardware module can offer a high-valued attestation (digital signature)on data, traceable to module’s security certificationUser opts to share data with app environmentApproach 2: Infrastructure (caution – privacy) Cell tower geolocation servicesCrowdsourced? (bitcoin block chain)#RSAC31 Justin Smith / Wikimedia Commons, CC-By-SA-3.0
Data at RestData in TransitTime and PlaceEndpoint Security
Trust means ?DB Independent security certification Key integrity Auditability / traceability Strong device identity credentials Robust application sandboxing System reliability Secure UI Data integrityDBAppserverUIAppserverAppAppOS / TrustZone /HypervisorHardwareSecure Key / IDstore managerCrypto#RSAC33
Apps require a secure, reliable foundation What gets to run on the platform? Am I in the real world or the matrix? Boot / code authenticationSecure debug lockEnvironment attestationPeripheral authenticationUI AppOS / TrustZone /HypervisorHardwareDo my secrets remain opaque? AppApplication partitioningHardware-based secure key storageSecure Key / IDstore managerCrypto#RSAC34
Example: Key protection Devices using secret or private key cryptography mustprotect their secret keysKinputsf()Device computes with key K, butbad guys must never extract KoutputsAttackers should not get K, even if they use mathematics,invasive attacks, external monitoring #RSAC35
Example: EM analysis of an RSA implementation Android app with RSA implementation onmodern 4G phoneMagnetic field pickup coilMeasurements collected duringcomputation of Md mod Nd 011S S M S M S01S M100S M SSStandards requiring sidechannel resistance PCI Movie Labs FIPS 140-3 Common CriteriaCF 36.99 MHz Acq BW 500 KHz Filt BW 250 KHz Smoothing 10#RSAC36
Trust from the top downDB DBDevice enrollmentAppserverSystem auditing & riskmanagementAppserverOnline revocationUIRemote management& updatesAppAppOS / TrustZone /HypervisorHardwareSecure Key / IDstore manager37Crypto#RSAC
Lifecycle considerations for “Internet Things”Limited UI foradministration steps“Direct to field”DeviceManufactureDeploymentEarly provisioning of dev. credentials Inject keys, certificates Enroll device May be done before OS load DeviceEnrollmentActiveUser/DomainChangeDevice administration secured by basecredentials Often an outsourced (faraway)manufacturing siteIn-field challenge/responseauthentication Add/update user credentials Send signed updates#RSAC38
Trust meets in the middleDeviceMfg.ServerDBIdentity key provisioningAuthentication serviceSecure session managementSecurity updatesDBAppserverUIAppserverAppAppOS / TrustZone /HypervisorIdentity key managementSandboxed secretsPartitioning of critical stateReliability & integrityHardwareSecure Key / IDstore managerCrypto#RSAC39
Data at RestData in TransitTime and PlaceEndpoint Security
What’s next? The human Internet is a success story Yay for standards: TCP/IP, IETF, Apache, SSL/TLS But security has always played catch up!The Internet of Things is still the wild west Largely without security Proprietary and not interoperable And mashups always bring security challenges!#RSAC41
Internet We have many building blocks to securethe Internet of ThingsBut they must be applied to solve adifferent (and changing) set ofchallenges!Think carefully before you buildtomorrow’s legacy problems!#RSAC42
Questions?Benjamin JunChief Technology Officerben@cryptography.com#RSAC43
Identity key provisioning Authentication service App Secure session management Security updates Identity key management Sandboxed secrets Partitioning of critical state Reliability & integrity Trust meets in the middle Hardware OS / TrustZone / Hypervisor DB Secure Key / ID store manager Crypto AppUI App Device Mfg. Server App server server
