DARPA Hopes Mobile Carrier Will Adopt Secure 5G
A new program aims to secure the Internet of Things.
If all goes as planned, a major mobile cellphone carrier will ultimately adopt technology developed under the Defense Advanced Research Project’s Agency’s Open, Programmable, Secure 5G program. Doing so will allow the open-source, secure technology to proliferate as so-called Internet of Things technologies become more ubiquitous.
The program, known as OPS-5G, is developing a portable, standards-compliant network stack for fifth-generation (5G) mobile communication that is open source and secure by design. OPS-5G seeks to create open-source software and systems that enable secure 5G and subsequent mobile networks such as 6G, according to the program website.
Next-generation wireless technologies are expected to fuel a dramatic rise in the number of networked devices commonly referred to as the Internet of Things, or IoT. The IoT could lead to the development of smart homes, smart cities, smart military bases and potentially even smart battlefields.
The signature security advantage of open-source software is increased code visibility, meaning that code can be examined, analyzed and audited, either manually or with automated tools. In addition, the portability of open source serves as a desired side-effect, to decouple the hardware and software ecosystems. This significantly raises the difficulty of a supply-chain attack and eases the introduction of innovative hardware into the market. The program seeks to enable a “plug-and-play” approach to various software components, which reduces reliance on untrusted technology sources, the website explains.
Programmable networks bring new flexibility to 5G but also introduce novel security challenges. To achieve 5G’s potential benefits, programmability must be implemented and managed carefully. Programmability must also be developed in ways that avoid rampant opportunities for misuse. OPS-5G aims to improve overall 5G security by increasing trust at a set of soft points that include unmanaged, unattended, long-lived and possibly long-forgotten IoT devices. Additionally, OPS-5G aims to address unintended and unwanted interactions between network slices and threats from the vast increases in network scale.
The OPS-5G program manager has set an unusual goal for the third and final phase of the program: to have one of the major mobile carriers adopt the technology by the end of the effort. “That’s a very weird criteria, but I think that unless you can see a carrier near a military base, see this as being a viable option, the game is lost,” admits Jonathan Smith, OPS-5G program manager within the agency commonly known as DARPA. “That’s a different way of doing business, and that’s, I think, a very innovative feature of this program. It’s intended to produce deployable pieces by the end game.”
While the ambitious goal poses a challenge, Smith says there are reasons to hope for success. “We’re building on top of a platform, the Linux Foundation’s software platform, that is already used by carriers and supported by them to some degree, so part of the problem of achieving entry … into mobile network operator networks is eased by basically hitching ourselves onto a platform that’s already achieved some of that,” Smith notes.
The four-year program officially began last fall with a so-called kick-off meeting with representatives from companies that had responded to DARPA’s broad agency announcement released in January of last year. The first two phases will be 18 months long, the final phase one year. It also covers four technical areas: machine translation of 5G standards; security and size, weight and power; secure network slices, which essentially allow creation of multiple virtual networks on a single physical network; and programmable defenses.
Machine translation of open source standards may be the most challenging technical area. While many humans may enjoy reading classic novels, such as Finnegans Wake by James Joyce, it probably could never be boiled down into computer code, Smith notes. On the other hand, technical standards that may include flow charts illustrating the interaction between nodes in a system, along with corresponding text and specific parameters, are extraordinarily tedious for humans to read but may be a good fit for computers.
“There seems to be the possibility that you could essentially use some of the techniques that have been developed in the natural language processing community to try to analyze the standards documents, which are very regimented and rigorous, and turn them into code,” Smith suggests. “We got some brilliant proposals in that area, and so we have a moonshot in there in [technical area one], which is that we’re trying to get the performers to build software that will read the standards documents and produce at least material in an intermediate form.”
Doing so would accelerate code production, especially since “software is the enemy of most deadlines,” as Smith puts it. He describes in layman’s terms how the process might work if, for example, a new standards document is published. “We suck in the new standards documents, and this gives us a head start, and even most of the work is taken over by software, which reads the standards and helps the programmer maybe to the point of getting rid of almost all of the work except some small modifications to get to working code,” he elaborates. “That’s probably the craziest thing that you could imagine, but people are very hopeful that it will make a difference.”
He expresses satisfaction with the teams working that part of the program. “I’ve got two excellent teams working on that, one from SRI and one from CACI.” SRI International is an independent, nonprofit institute, and CACI International Incorporated is a nearly $6 billion national security and defense company.
Regarding security, DARPA intends to incorporate the zero trust architecture, which essentially requires authentication and verification of any user or device seeking to connect to a network. “The idea is that you are not promiscuously keeping all of your services accessible all of the time. That reduces the attack surface both in time and space. It’s a very elegant idea. We have projects underway in that area with MIT, Perspecta and Kryptowire,” Smith reveals.
The program manager compares network virtualization to data center virtualization that has taken place in recent years, with companies such as Amazon and Microsoft developing large-scale, professionally operated clouds computing centers, virtualizing computer power and storage. “So, 5G is essentially giving you—with slicing—the ability to define a virtual network of the flavor that you want. You will define a network of services, and this is something that has been enabled by new programmable elements, such as software-defined network switches and network function virtualization, NFV,” Smith says.
The capability allows networks to be more finely tailored to user needs. The Defense Department’s needs may be very different from those of a warehouse or hospital, for example. Smith describes two robotic arms picking up a vat of liquid in a warehouse. They would need to be highly synchronized, which may involve selecting pathways through the network requiring “bounded delays through each network element and minimal contention at the network elements,” and those bounded delays are “something you might want to customize.”
An aerial sensor collecting data over a given area and transmitting it to military analysts, however, would not need a bounded delay but may need to use all possible paths through the network fabric. “You can imagine a slice that would be defined as having multipaths and multipath forwarding, so delay is less of an issue, but very high throughput and very high response time to get the sensor pull back to the U.S. would be the priority. So, you’d set it up very differently because you’d want your slice configured to utilize the underlying network resources in different ways to achieve different objectives,” he says.
Unfortunately, that promising capability also offers risks for the Defense Department. Smith cites a theoretical example of a Defense Department organization having to share technology with an unknown entity, somebody who is not a friend.
“There have been a lot of very interesting technical attacks that have been done on platforms with shared resources where basically somebody on a shared resource, such as a shared switch or shared server, has been able to pull your cryptographic keys and what you’re doing essentially by looking at the shared hardware resources such as caches on the system,” he points out. “What we are doing in the secure slices technical area is looking at ways in which we can designate or select a particular slice, a virtual chunk of the network and designate it as being secure and then insulate it so that there’s no information flow to potential adversaries.”
He hints at additional security capabilities. “We have other security features that are designated in there as well that the researchers are working on, and we have a performer there, which is USC’s Information Sciences Institute, usually called ISI, and they have some very cool ideas,” Smith says, referring to the University of Southern California.
The DARPA plan is to use programmability in defense of the network. “We’re trying to harness the programmability as a way to rapidly adapt to the defense of the network so that we could load software into the boxes and detect that something like this was happening and do one or more algorithms, such as DDoS [distributed denial of service] pushback that have existed only in the literature, but with OPS-5G, I think can be rapidly put in place to defend the 5G core,” Smith says. “What we’re trying to do is turn the programmability around to be a force for good rather than a risk factor.”