Identity, Security and Privacy

in the IoT era

Gregory Estrade - http://lvt.tl/

@Torlus on Twitter, GitHub.

IoT - State of the art

Focus of this talk

Home automation using MQTT

A quick MQTT overview

  • MQTT is based on a publisher/subscriber model.
  • MQTT clients use TCP/IP to connect to a Broker.
  • Brokers perform message-passing between clients and can usually be chained in a hierarchical way.
  • There is a MQTT-SN specification for low-end devices.

Home automation using MQTT

Architecture basics

  • Sensors: room/temperature, outside/humidity.
  • Commands: room/radiator, garage/door.
  • All devices are using a MQTT broker connected to Internet.

Privacy requirements

  • Some data may be shared publicly: outside/humidity.
  • Some data may be shared to a trusted set of people or systems: garage/door, room/temperature.
  • Some commands need to be protected: room/radiator.
  • Some commands need to be secure: garage/door. A 2FA would be a good thing.

Home automation using MQTT

Connecting our home to the IoT

  • Naming: my_home.domain.name/<location>/<device>.
  • Security: TLS required for MQTT requests from Internet.

Well, it was easy, wasn't it?

End of the story?

.NOPE.

DNS and SSL/TLS concerns

Threats on DNS

Issues with SSL/TLS

Both share the same usage flaw.

 

Hierarchical

Trust Model

Enter the Blockchain

The Blockchain

  • Made famous by cryptocurrencies, especially Bitcoin.
  • In cryptocurrencies, acts as a ledger for transactions.
  • Integrity is ensured by the network. No SPOF.

Why does it work?

  • The mining process rewards people for maintaining the ledger's integrity. Mining is computational power-hungry.
  • In cryptocurrencies, everyone in the network is motivated by the same incentive, i.e. getting money.
  • It works because everyone is a bit greedy and that the "silent majority" outperforms the "few villains".

It looks like a very negative view of human nature, but...

TRUST < GREED

History has proven that relying on the "good will" of a few ones, just doesn't work.

Decentralization matters

Yes, all that stuff looks interesting, in theory at least...

But who cares ?

January 2015 - IBM and Samsung unveil ADEPT

Autonomous Decentralized Peer-to-Peer Telemetry
Using a Blockchain with Ethereum, Telehash for communication, and BitTorrent.

However, nearly 8 months ago... Mosquitto-Twister

Proof-of-concept of a fully distributed MQTT broker for the IoT. Uses Twister platform, Bitcoin and Bittorrent protocols. https://github.com/Torlus/mosquitto-twister

How does it work?

The Twister platorm

Twister is a fully decentralized Twitter clone.

  • A Blockchain stores the full list of User Handles.
  • User profiles are stored in a DHT (Kademlia).
  • Posts are stored within BitTorrent swarms.
  • DMs are encrypted, other messages are not.

Mapping into MQTT concepts

...and addressing security and privacy as well, while we're at it...

  • A Handle is used instead of a Domain Name.
  • MQTT's pub/sub model fits nicely with Twi[st]ter's one.

Now what?

Blockchains and incentive issues

  • Generally speaking, what would be good incentives for non-cryptocurrencies-related Blockchains?

    Example: DNSChain.

  • Blockchains quickly become huge.

    Would a distributed Blockchain storage be possible or relevant? What about integrity? Use another Blockchain for this purpose ?

Other subjects

  • Cryptography and low-end devices.

    Cryptography is costly. Do we need strong cryptography, when EFT standards still use DES (2-TDES, 112-bit key length) ? Hint: DUKPT.

  • Most cryptography algorithms rely on the difficulty of performing some tasks.
  • http://en.wikipedia.org/wiki/One-time_pad is very simple to understand, yet:
    • It is one of the few systems proven unbreakable (Shannon, 1949)
    • It has been invented by Frank Miller, a banker.

Other subjects

  • Secret sharing.

    Secret keys are hard to protect. Storing a key into a device makes it a SPOF, so let's split the key into parts and store those parts across multiple devices... Hint: ssss.

  • Adaptive security.

    NFC payments are prone to errors, but limited to a maximum amount. It's a matter of risk versus consequences. Let's extend this principle to give a proper response to a given situation, using a Quorum.

Questions?

Gregory Estrade - http://lvt.tl/

@Torlus on Twitter, GitHub.

Bonus: Security and Trust

Let's compare two different security strategies

Topic

Online services

Payment cards
Passwords 8+ characters alpha, num, special 4 digits PIN code
Cryptography Asymmetric Symmetric
Algorithms RSA, AES DES (2TDES)
Key sizes 2048, 4096 112 (2 x 56)

Failure Impact

 Embarrassing party pictures leakage Bankrupcy, for you and your whole family