First of all, the IoT is the absolute dream come true for eavesdropping and a complete loss of a person's privacy and anonymity. Fine dining anyone?
An open ecosystem which IOTA offers makes it a lot harder to pull something like this off, if only it's a big improvement over proprietary closed source IoT where things can be hidden. All IOTA software is open source and therefore less vulnurable to something like this. Using a public ledger doesn't mean your data can't be private. IOTA messages can be encrypted as you see fit; We even offer Masked Authenticated Messaging with a set of pre-defined ways to do this including Public Key Encryption support.
The ecosystem that your cryptocurrency hopes to operate in is an electrical swamp: a plethora of devices, little standardization, uncertain updates; significant power, storage, and processing limitations for security (authentication, confidentiality, non-repudiation, integrity checks).
The current IoT ecosystem is in its infancy. IOTA is trying to mature it by introducing standardization (together with for example the Object Management Group and by being a founding member of INATBA) and by providing a open, owner-less and permission-less protocol for IoT. Because the IoT we foresee isn't there yet we have a chance here to change the status quo.
So, how to secure it? Let's do the one thing that anyone with an iota of experience says not to do--let's come up with our own cryptographic solutions, our own implementation(s) of cryptographic hash(es), etc.
So "Don't roll your own crypto" means we can never create new cryptographic solutions again from now on just because? I don't think that's how it works moving forward. I think it's a valid point for individuals who want to re-invent the wheel, this is not the case here. IOTA uses Ternary instead of Binary which theoretically will give a big advantage over Binary based solutions once the hardware is available. Because IOTA uses Ternary the existing Binary based hashing solutions won't cut it. Therefor IOTA works together with the world leading cryptographers from Cyber-crypt to develop Ternary based cryptography like https://www.cyber-crypt.com/troika/. This is not just some "Roll your own crypto", this is the development of a new and open cryptographic hash function with a 200k EUR bounty program and extensive peer review, a similar route conventional cryptography takes to become a standard.
Shouldn't IOTA instead try to follow an already proven path for cryptographic services?
That might be a short-term shortcut, but we are working towards a standardization that will last for decades, this includes being Quantum immune and Ternary ready. We think Ternary will have a bright future once the hardware is there so we are preparing for that. This includes moving away from existing cryptographic solutions since they are simply not suitable for efficient Ternary hardware.