We are quite familiar with where IoT is heading. We mostly anticipate a positive outcome, and hence those anticipations drive us to find and do more towards IoT's development. One such promising idea is block-chain driven crypto-technology. Although it is in its infancy optimists have already invested a good fortune towards its development. For example, we have IOTA cryptocurrency being traded as we speak. In one hand we shall have machines to collect payments or share labor costs with one another. On the other hand, we shall have security challenges like trust.
What are your predictions on pros and cons of the machine-to-machine economy?
While it seems we are certainly heading into a world where machines will be autonomous and transact with one another without human intervention, I see a huge gap as concerns trust. I think there is need for further research into trust applications and dispute resolution between machines. Certainly if we put it in code, we will get better ecosystems than perhaps today's human dispute resolution which is prone to corruption and flawed judgement.
Looking at this machine to machine trust issue logically, it is not anything new. Not really.
How does crypto work today? In some environments, such as when people use a secure web connection from home, the secure connection ensures that the session the user has launched is indeed between that one user's computer and the web site. It makes intrusion into that session difficult, because only that one client and that one server can decrypt the two-way conversation. In many cases, the secure session includes authentication with username/password, in which case, security includes the identity of that one individual.
Okay, so ultimately, in these home scenarios, we have security built into a machine to machine interface already. In the case of also having to enter username/password, it is a client-server model, in which the server only allows access to a certain client, into that client's account.
In an enterprise network environment, the presumption is that PCs are each operated by one person. The first PC transmits its public key, allowing the other individual, the second PC in the conversation, to return an encrypted message that only the first PC can decipher. In doing so, the second PC sends its public key to the first PC, and now they can exchange encrypted packets in both directions. But only between these two users! The process has to repeat, for every other user with which you need secure communications. (And once encrypted communications are possible, the users can also exchange secret keys, to allow for faster symmetric key algorithms to be used.)
In this enterprise environment, what is the main issue? It is one of saying, I trust that this public key I have been sent is indeed from Joe Smith's computer. There is a process in place, mostly a manual administrative process, which clearly identifies the owner of that public key. That's why you get authentication and encryption certificates, with the identity of the owner of that certificate. Some manual process is established, to tie that user's identity to the computer he has been assigned.
So ultimately, here too, it's also a machine to machine interface that has been trusted.
For machine to machine interfaces that don't involve a human user at all, no reason to think the situation is drastically different. We need to bake an identity into each legitimate processor built into that car, let's say. So that the system integrator will allow only those ECUs that have the proper credentials to communicate with other ECUs. Once the system has been built, it will not be possible for an intruder to communicate with any ECU in that car, unless this new ECU's public keys have been accepted by the other ECUs, assuming we go initially with asymmetric key algorithms as we did in the enterprise network. That process should involve manual intervention, when the system is updated, or a component part changed. Or if we assume secret keys only, for only symmetric key algorithms, then it's possible to distribute the secret key(s) under certain controlled conditions. And perhaps update these on some regular basis.
In short, I don't think the security aspects are all that different, for IoT, other than the scale. Also, no reason to assume that all of what might be "IoT" needs to be very secure. Perhaps it doesn't matter greatly if that thermometer you have on your porch is not absolutely secure?