Schneier on Security

stimoceiver • December 29, 2015 11:24 AM

Its time to DECENTRALIZE THE INTERNET!

How long has the forthcoming Internet of Devices, the Internet of Things, been heralded? Years? Decades? And as of yet we’ve seen very little in the way of the actual implementation and deployment of standards that would enable such to come to fruition.

This article suggests one reason why. But I’d like to suggest another.

Clearly we couldn’t really achieve something like the Internet Of Things without something like IPv6. To attempt to do so using IPv4 would have been madness. Possible perhaps, using NAT and other technologies. But surely a pointless exercise in complexity.

And yet it is now over 3 years since World IPv6 launch day and still we find the industry, normally a juggernaut fully willing to impose planned obsolescence upon us all to achieve its ends, still moving at a snails pace when it comes to implementing the open standards and protocols the Internet of Things will require.

What kind of standards will we need? Ubiquitous adoption of IPv6 in our devices and software, our protocol stacks. An IPv6 variant on dynamic addressing capabilities of DHCP. Dynamic routing between peers. Roaming and at least the possibility of interoperability for guest devices. Some of these standards and protocols already exist. And yet they are sorely lacking in the kind of ubiquitous adoption in the real world that signals when a standard is really a standard.

And we have one of the Chief Figureheads of no less than the US Central Intelligence Agency, David Petraeus, waxing poetically about his lust to co-opt and commandeer the as-yet-forthcoming Internet of Things for purposes of spycraft:

“Items of interest will be located, identified, monitored, and remotely controlled through technologies such as radio-frequency identification, sensor networks, tiny embedded servers, and energy harvesters — all connected to the next-generation internet using abundant, low-cost, and high-power computing {…} .”

I’d like to suggest another reason we don’t yet have an Internet Of Things. And that is simply, that the evolution of standards required for an Internet Of Things are nearly identical to those we would need to build a truly open, node-centric or device-centric peer routed mesh. A new kind of Decentralized Internet where all one has to do is acquire a radio and put up an antenna and start networking with ones immediate neighbors. Where the only costs of access are the cost of the WAN digital radio and the antenna – and the “costs” of routing traffic for ones neighbors.

Sure it would be slower – at first. But thanks to IPv6 being fully a superset of IPv4, interoperability between such a new, free, peer routed mesh and the current style of proprietary, “tethered to the infrastructure” internet access most of us enjoy, is guaranteed.

We’ve become so immersed in a networking paradigm of isolated WLAN islands and multiple layers of firewall that we fail to see just how long ago we could have instead OPENED our networks. I understand the security concerns that led the development of firewalls. And I recognize that at least here in the big US city where I live, neighbors rarely talk to neighbors. But now that its the 21st century, hindsight shows we clearly could have had far greater interconnectivity between adjacent nodes years, perhaps decades ago, without sacrificing any more in the way of security than we do now when we forward a port through a NAT firewall. Or when UPNP does it for us.

The time has clearly come to retool the internet to be fully free and fully decentralized. Sure commercial paid ISP access will still exist. But imagine the possibilities for civic scale networking and organizing once we are able to link up entire neighborhoods in a single network. Not-for-profits, hackerspaces, businesses, and tech-savvy consumers could volunteer to establish the backhaul connections.

And the beauty of doing all of this in IPv6 address space is, the network doesn’t NEED to exist physically before it can exist logically. In the early stages of adoption when free backhauls are scarce, one could simply connect geographically disparate neighborhood networks through the existing IPv4 network. Something like Teredo in reverse. Future proofing for encryption and anonymity via darknet style routing can be built right in to the protocol stack, but fundamentally this is about creating a streamlined network out of commodity consumer hardware, so most likely the “Layer 3” protocols would be fully open, at least in the beginning, and we could still rely on existing methods such as SSL for encryption.

And since the network is totally free and open, why not build in dual naming for nodes? Those who desire a semi-permanent namespace can use a certificate based authentication of nodes on the same decentralized scale as the backhauls. Think of how the old .us TLD was broken down geographically from most-significant to least-significant. USENET’s style of administering the Big-7 groups also comes to mind, although in that case the most-significant and least-significant parts of the namespace are reversed. Names could be cryptographically hashed to radio MAC addresses. This could even allow for roaming.

But that is only one style of naming. USENET also has the “Alt” hierarchy with its first-come, first-serve naming and so it might even be possible for the new network to also embrace this level of freedom, at least for geographically fixed nodes.

I can only speculate as to reasons that some of these advances haven’t already happened. Think of just how long Windows had a “Network Neighborhood” right there on Explorer. Due to security fears this was historically used more or less solely in corporate domains and workgroup environments.

But surely open standards could long ago have just as easily been designed for the “Network Neighborhood” to provide user to user messaging, email, forums, content syndication (blog/wall), tiered or access-list based privacy, file transfer, and even something like VLANs for gaming purposes, on the early wired networks – and again, later, on early wireless networks. (I know some of these things did exist – but never in a ubiquitous standard embraced in commodity consumer software and hardware.)

In my opinion, now that the “Internet Of Things” is entering the mainstream consciousness of a far more digitally savvy populace, now that industry is once again making so public a push to develop these technologies that even the CIA is salivating at the possibilities, it is clear to me that there are many, many ethically and globally minded hackers who will not let this opportunity pass: To use the exact same new standards of the Internet of Things to turn the tables on the proprietary commercial internet we now have, and create a truly decentralized network of global scale yet neighborhood and civic relevance. A network far better at routing around surveillance and censorship. A network architecture that encourages interaction between neighbors instead of stifling it.

Its equally clear that the same points Mr. Schneier makes in this article apply to this new technology – and then some. All-inclusive ideas like this are among the biggest threats to the proprietary and the divisive. I wonder – could this be the reason the Internet of Things is taking so long? Could repressive regimes – both governmental and otherwise – have already envisioned this long ago, and carefully steered technology away from such innovations that enable so potentially vast of individual, social and political empowerment and interconnectivity? I am reminded of Eben Moglen’s 2012 keynote presentation “Innovation under Austerity” at the Freedom to Connect conference where he gives an anecdote about government interest in thwarting anonymity (in addition to encryption) way back when he was defending Phil Zimmerman and PGP.

This may be our last chance to turn the internet into the free and open civic network it could have been since the beginning.

stimoceiver • January 1, 2016 7:55 PM

@V its funny you mention Fidonet. I dont know much about their underlying architecture but IIRC it was a store-and-forward system like USENET.

More relevant from that era is the transition from bang path notation for email and news to using DNS.

I merely point out a gap in the existing implementation of “neworking” in our daily lives. Here are some others. Why dont our smartphones have the ability to recognize each other, say, on public transportation? It would enable people whose privacy settings include a public facing profile to game or strike up conversation with each other based on mutual interests. Why don’t our modern, digitally networked cars include the ability to communicate with other nearby cars? One could go back far enough and ask the same question of television: Why has it remained only a “one-to-many” communications medium for so long?

What consumers dont understand is that we have become so dependent upon infrastructure that we neither own nor control that we put ourselves in a dangerous position. When an entire network of intercommunicating devices exists between multiple devices in your household, over distances measurable in inches and centimeters, some routing traffic for others, the resulting topology may reasonably be termed a mesh. And how much of this mesh will include devices like smart televisions and smartphones, devices with cameras, microphones, and network connections that can be used to surveil? devices that are closed-source and almost always laden with inbuilt vulnerabilities? Especially devices like smartphones where unbeknownst to most, the baseband architecture is sophisticated enough to be running its own, proprietary second operating system administered via infrastructure-facing interfaces only, and already known to be fully capable of doing things like updating the end-user-facing commodity OS or activating the microphone. Today’s smartphone vendors could end up being the first vendors of mass market wearable and implantable computing. Do we really want those devices to share the design philosophy of a proprietary baseband architecture and OS, physically and cryptographically secured against end-user rooting and modification?

These same or similar critiques apply to the network architecture itself. Why shouldn’t we radically reorient our networking paradigms to include node-centric (mesh) networking? Allowing ONLY proprietary-routed networking ensures the infrastructure-owners will always have the upper hand when it comes to surveillance and data mining.