New Cyber Standards for IoT Ease – But Won’t Solve – Security Challenge
The first independent standard for cybersecurity for the Internet of Things (IoT) was approved earlier this month, following two years of debate and discussion over how to measure and secure such devices.
The American National Standards Institute (ANSI) approved UL 2900-1, General Requirements for Software Cybersecurity for Network-Connectable Products, as a standard on July 5. The effort was spearheaded by Underwriters Laboratories (UL), which is preparing two more standards to follow: UL 2900-2-1, which defines requirements for network-connectable components of healthcare systems, and UL 2900-2-2, which does the same for industrial control systems.
The three establish the first standard security protocols for software-controlled IoT devices, such as access controls, industrial controls for lighting and mechanical systems, internet-connected medical devices and more. They also offer potential answers to major worries about the lack of security built into such devices thus far.
The Internet of Things promises unparalleled opportunities to track, control and manage everything from lights and security cameras to pacemakers and medication delivery systems. But concerns about security, driven by real-world events in which unsecured IoT devices were co-opted in coordinated botnet attacks, have raised anxiety levels about the risks posed by connecting so many devices to the internet.
Those concerns have prompted government leaders from the Pentagon to Congress to call on industry to embrace security standards as a mark of quality and establish voluntary independent testing programs to assure customers that products are safe. The underlying warning: Either industry figures out how to police itself or government regulators will step in to fill the void.
Whether that is enough to inspire more companies to step up to the standards challenge remains unclear. “The market – that is, individual, corporate and government customers – has yet to put a price on IoT security in the same way that other markets have to determine the relative value of energy-efficient appliances or crash-worthy automobiles,” said Chris Turner, solutions architect with systems integrator General Dynamics Information Technology. “The market would benefit from standards. They’d help vendors back up product claims and integrators speed up adoption and implementation, which in turn would increase security and probably drive down prices, as well.”
A standards regimen could change that equation, suggests Steven Walker, acting director of the Defense Advanced Research Agency (DARPA).
“What if customers were made aware of unsecure products and the companies that made them?” he asked at the AFCEA Defensive Cyber Symposium in June. “I’m pretty sure customers would buy the more secure products.”
As recently as Oct. 21, 2016, the Mirai botnet attack crippled Internet services provider Dyn via an international network of security cameras that launched an onslaught of bogus data requests on Dyn servers, peaking at about 1.2 terabytes/s. The attack brought down many of the most popular sites on the Internet.
Kevin Fu, director of the Archimedes Center for Medical Device Security and the Security and Privacy Research Group at the University of Michigan and the co-founder and chief scientist at Virta Labs, a startup medical device security firm, told the House Energy and Commerce Committee that the underlying problem is one of market failure.
“We are in this sorry and deteriorating state because there is almost no cost to a manufacturer for deploying [IoT] products with poor security to consumers,” he said at a November hearing. “Has a consensus body or federal agency issued a meaningful IoT security standard? Not yet. Is there a national testing lab to verify and assess the pre-market security of IoT devices? No. Is there a tangible cost to any company that puts an insecure IoT device into the market? I don’t think so.”
Could UL 2900 answer that need? Though Fu isn’t quite ready to endorse it, he did suggest the concept is sound.
“We know from the mathematician Gödel that it’s impossible to have both a sound and complete set of standards for any non-trivial problem,” Fu told GovTechWorks. “However, standards are important to improve security and simplify the problem to make it more tractable. No approach will completely solve security, but standards, sound engineering principles and experience gained through failure are necessary ingredients for reasonable defense.”
Developing the Standard
UL 2900 provides guidelines for how to evaluate and test connected products, including a standard approach to software analysis, efforts to root out embedded malware and process and control requirements for establishing IoT security risk controls in the architecture, design and long-term risk management of the product.
Rather than focus on hardware devices first, UL focused on software after initial conversations with the Department of Homeland Security (DHS), said Ken Modeste, leader of cybersecurity services, at UL. “One of DHS’s biggest challenges was their software supply chain,” he said. DHS was concerned about commercial software products running on computer systems, as well as industrial control software running the agencies operations technology, such as air conditioning, lighting and building or campus security systems.
Examining the problem, UL officials found clear similarities between the systems and sensors used in factory automation, enterprise building automation and security technology. “The majority of these cyber concerns – 90 percent – were in software,” Modeste told GovTechWorks. “So we realized, if we can create a standard for software, we can apply that to many, many products.”
UL invited representatives from industry, government and academia to participate in developing the standard. “We started looking at industry standards that make software better,” Modeste said. “A firmware file has a multitude of components. How can those be broken down and understood? How can they be protected?”
Participants studied every imaginable attack vector that threat actors could use to compromise a product, and then incorporated each into the testing process. Recognizing that new threats and vulnerabilities arise all the time, the testing and process was designed to be fluid and to incorporate follow-up testing after initial approval.
At first, Industry was slow to respond. “I thought we’d have more support early on,” Modeste said. “But there was an initial reluctance. It took a while for us to engage and get them to see the advantages.”
Now it seems interest is on the rise. Among the first movers with the standard: Electric Imp, an IoT software firm based in Los Altos, Calif., and Cambridge, U.K., which provides a cloud-based industrial IoT platform for fully integrating hardware, operating system, APIs, cloud services and security in a single flexible, scalable package. The Electric Imp platform is the first IoT platform to be independently certified to UL 2900-2-2.
Hugo Fiennes, co-founder and CEO at Electric Imp and former leader of Apple’s iPhone hardware development efforts (generations one through four), said:
“For security, UL has come at it at the right angle, because they’re not prescriptive,” Fiennes told GovTechWorks. “There are many ways to get security, depending on the application’s demands, latency requirements, data throughput requirements and everything like that. [But] the big problem has been that there has been no stake in the ground so far, nothing that says, ‘this is a reasonable level of security that shows a reasonable level of due diligence has been performed by the vendor.’”
What UL did was to study the problems of industrial control systems, look at the art of the possible, and then codify that in a standard established by a recognizable, independent third-party organization.
“It can’t be overstated how important that is,” Fiennes said. UL derives its trust from the fact that it is independent of other market players and forces.
Although UL 2900 “is not the be all and end all last word on cybersecurity for IoT,” Fiennes said, “it provides a good initial step for vendors.”
“They haven’t said this is one standard forever, because that’s not how security works,” he said. “They’ve said IoT security is a moving target, here is the current standard. We will test to it, we’ll give you a certificate and then you will retest and maintain compliance after.” The certification lasts a year, after which new and emerging threats must be considered in addition to those tested previously.
“This doesn’t absolve the people selling security products, platforms and security stacks from due diligence,” Fiennes warned. Firms must be vigilant and remain ready and able to react quickly to threats. “But it’s better than nothing. And we were in a state before where there was nothing.” He noted that his product’s UL certification expires after a year, at which point some requirements are likely to change and the certification will have to be renewed.
Still, for customers seeking proof that a product has met a minimum baseline, this is the only option short of devoting extensive in-house resources to thoroughly test products on their own. Few have such resources.
“Auto makers and other large-scale manufacturers can afford that kind of testing because they can spread the cost out across unit sales in the hundreds of thousands,” says GDIT’s Turner. “But for government integration projects, individually testing every possible IoT product is cost-prohibitive. It’s just not practical. So reputable third-party testing could really help speed up adoption of these new technologies and the benefits they bring.”
Standards have value because they provide a baseline measure of confidence.
For Electric Imp, being able to tell customers that UL examined its source code, ran static analysis, performed fuzz testing and penetration testing and examined all of its quality and design controls, has made a difference.
For UL and Modeste, the notion that it will not be able to solve the IoT security problem with a single standard, proved something of an “aha moment.”
“Within cybersecurity, you have to recognize you can’t do everything at once,” he said. “You need a foundation, and then you can go in and take it step-by-step. Nothing anyone comes up with in one step will make you 100 percent cyber secure. It might take 10 years to come up with something perfect and then soon after, it will be obsolete. So it’s better to go in steps,” Modeste added. “That will make us increasingly secure over time.”