The government is justifiably concerned about the existential vulnerability that has arisen because of the connection of infrastructure controls, i.e., supervisory control and data acquisition (SCADA), to the public networks. This connection permits at least parts of the infrastructure to be operated from any place in the world. To the extent that the controls are insecure, they can be abused or misused to cause the infrastructure to be mis-operated.
To the extent that the infrastructure itself is fragile, mis-operating it may cause damage that cannot be efficiently remedied. "Experts" have speculated that the infrastructure might be maliciously operated in such a way as to shut down our entire economy for days to weeks. Value and savings might be destroyed. Millions might starve or freeze unless we could rebuild in days what it has taken us decades to create.
However unlikely such an event, to the extent that such a vulnerability is implementation induced, rather than fundamental, it should not be tolerated. Making controls intended only for the use of a few privileged operators visible to everyone is unnecessary, in this case, reckless. It is analogous to putting a copy of the control of the autopilot for an airliner between every two seats.
However, it is specifically because the infrastructure is fragile that the controls are connected to the public networks in the first place. The operators understand that the infrastructure must be "operated;" that its continued service requires that it be monitored, adjusted, "provisioned," and configured to compensate for changes in inputs or load or the inevitable failure of components. While some of this operation is automated, some of it requires timely human intervention.
The operators of these controls have connected them to the public networks on the implicit assumption that, far bigger than the risk of connecting them would be their own inability to monitor and operate the controls on a timely basis. Few of them see their connection in the context of all the other connections. They understand that no single connection would represent a major risk; they are only just waking up to the realization that the collection constitutes an existential vulnerability.
Part of the problem in the Critical Infrastructure space is the culture. Given the sensitivity of these controls, one would expect them to be hidden behind virtual private networks and strong authentication. For reasons of convenience, for the most part they are not and that is the root of the problem.
In order to provide for around the clock, but somewhat sparse, remote monitoring and control, the operators have have connected the controls to, not just one, but to both of the public networks. While this kind of remote operation is good for the enterprise, and may even be strategic, many of the early connections were tactical, more by and for the convenience of the operators than for the enterprise.
In order to improve the chances that they can always connect when necessary, that is, compensate for any network failure, many of the controls are connected both to the public switched telephone network (PSTN) and the Internet. While they use the public wide area networks, they use them to create a limited number of relatively short point-to-point connections, for example, from the operator's home to the plant.
While the public networks permit world-wide any-to-any connectivity, and while the operators might actually monitor and operate their systems from the end of a plane trip, that is the exception, not the rule. The result is that anyone may use the public networks to send a message to any of these controls. They may be able to connect and operate the controls.
In the early days, most of these controls were purpose-built, offered only a limited command interface, and operating them required a lot of special knowledge. Even finding them would have been difficult, much less misusing them. Today, many have already been identified; most of them have graphical user interfaces and require much less special knowledge. Moreover, such intelligence as operator manuals and other documentation may be independently available in the world-wide-web.
Behind the controls, there may be operational dependencies between components such that operation of one may influence the behavior of others. For example shutting down the external power to a nuclear reactor may cause the reactor to shut down. These effects may cascade. The electrical grid is the most inter-dependent of the infrastructures and almost everything else is dependent upon it.
What might a "separate" network look like? How separate might it be? Well, it might be as separate as the two public networks are from one another. For example, it might have a separate "address space." Like these two networks, it might use different signaling, connection setup, and protocols.
On the other hand, the Internet, the digital network, originally piggy-backed for connectivity on the PSTN, the analog network. Today, for reasons of efficiency, all wide area networks share the same glass and copper fabric and most analog traffic is now encapsulated in digital. While much of that fabric is less than a decade old, it has taken us more than a century to achieve near world wide coverage. Surely a new separate network would exploit the existing fabric rather than attempt to replicate it.
For security reasons, it might be desirable for the networks to have different user populations. However, that would mean that a user of the alternate network could not use the public one. Not very likely.
The single public fabric that we use today emerged as a number of public and private networks coalesced around the Arpanet. When I first became an e-mail user, I had a list of tens of gateways and paths from the IBM network to other networks. We would use nested addresses of the form ((foo@foonet)@ibmgatewaytofoonet.com). Sometimes these addresses were two or three layers deep. An x400 or proprietary address might be nested inside an IP address or vice versa. Routing through these gateways often required a great deal of special knowledge. Gradually those gateways gave way to intelligent routing. x400 and other forms of addressing gave way to IP addressing.
The Internet is defined, and has evolved, as the collection of all networks that are connected to one another, that communicate in Internet protocols, or that are connected via gateways, think firewalls, that use that protocol. We did not set out to have one network; there was no design or intent. The Internet came about for economic reasons. The value of a network goes up with the number of potential connections. Therefore, the propensity of two networks to connect goes up with the square of their size. The unfortunate corollary to this is that, if we were able to provide a separate network, the users would respond to the economics by connecting them together again.
So for a number of cultural, technological, and economic reasons, a completely separate "alternate" network, no matter how desirable, seems unlikely. While still unlikely, a more viable alternative might be one or more virtual private networks (VPNs) exploiting the underlying fabric of the public networks.
Moreover. most of the advantages of such a network or networks can be achieved with much cheaper alternate mechanisms such as strong authentication, end-to-end encryption, and firewalls and other proxies. Even if there were hope for the kind of alternate network envisioned by Director Henry, it would still be our job to apply those mechanisms while we were waiting for it to emerge. It is also necessary to hide all of the information about the infrastructure controls that is gratuitously available to all but needed only by the few.
The status quo is the result of a large number of individual but reversible choices. It is unacceptable. It is our job to fix it. For that that we are called professionals and are paid the big bucks.