Creating A Roadmap For Hardware Security

The U.S. Department of Defense and private industry consortiums are developing comprehensive and cohesive cybersecurity plans that will serve as blueprints for military, industrial and commercial systems.

What is particularly noteworthy in all of these efforts is the focus on semiconductors. While software can be patched, vulnerabilities such as Spectre, Meltdown and Foreshadow need to be dealt with in hardware. In the past, efforts focused primarily on securing a supply chain for all technology and building an impenetrable firewall around sensitive data. Neither of those approaches is considered sufficient anymore.

The DoD’s current effort, being spearheaded by DARPA, is called the Automatic Implementation of Secure Software (AISS). “With AISS, we care about four attack surfaces,” said Serge Leef, program manager in the Microsystems Technology Office of DARPA. “It’s supply chain, side-channel attacks, reverse engineering and malicious hardware. The goal is to make intelligent tradeoffs involving the cost of security.”

Source: DARPA

Those tradeoffs can be significant. Rather than just storing authentication keys in a segment of a chip, to provide secure boot-up, there is widespread agreement that security needs to be multi-layered and included as part of the initial architecture. But some aspects of security need to be active, as well, which can add significant overhead to designs.

“The security mechanisms cost area, power and could impede performance, so you need to make intelligent decisions in which the user provides the cost function,” said Leef.

“This (roadmap) is intended to be everywhere. The DoD is just one beneficiary. Automotive is another. It includes everything from SoCs to ASICs, structured ASICs and FPGAs.”

DARPA has a long history of working with Silicon Valley. In fact, the Internet was based on a DARPA packet-switching network called ARPANET (Advanced Research Projects Agency Network), which was shut down in 1990. But over the past couple years, DARPA’s relationship with the chip world has really started to expand. It’s not unusual to see generals and other high-ranking officers sitting in presentations and attending semiconductor conferences these days.

Industry efforts
The U.S. government is far from alone in this effort. Global industry has a stake in tightening security, as well. Arm introduced its Platform Security Architecture in 2017, which it describes as a holistic set of threat models, security analysis, hardware and firmware architecture specs, and an open-source firmware reference implementation. The company has been issuing updates to its PSA ever since.

Siemens, meanwhile, has been following a similar timetable with its governance-and-consortium oriented approach, called the Charter of Trust, which it announced in 2017 and recently updated. The consortium will require OEMs that provide particularly sensitive hardware or software to follow specific security protocols, and eventually do security reviews themselves rather than relying on Siemens. The company is pushing other members of the group, which include Dell, IBM, Cisco, Daimler, NXP SGS, Deutsche Telekom, and TÜV SÜD, to enforce similar rules on their OEMs to create a more consistent set of expectations for all tech providers.

Being able to identify a device for certain and track it back to its origin reduces the risk that it could have been diverted long enough to be compromised.

“If it’s a connected device—and there aren’t too many things anymore that just sit on your kitchen table—you want a unique identity key that would let you chase it through the supply chain, know you could identify it in a way that can’t be tempered with,” said Michael Chen, design for security director for the New Ventures Division of Mentor, a Siemens Business. “That requires cryptography, which requires public and private keys, which means having a system to manage the keys and certificates that you know are valid. You really need a multilayered approach you can keep up through the lifecycle of the device.”

Supply chain
The U.S. DoD has been pushing toward much tighter control of its supply chain as the threat of cybersecurity attacks grows, particularly from China and Russia, and industry has been following closely behind.

“The supply chain integrity question is increasingly important because the limited number of companies that can handle a chip throughout its production cycle and the fragmented supply chain means having to hand parts of the process off to a series of unknown or untrusted entities,” said Ben Levine, senior director of product management for the Rambus Security Division. “That leaves a lot more opportunity for gray-market production, introduction of security vulnerabilities, cloning of devices, and it puts companies in a position of having to improvise security for places in the supply chain we don’t control and may not have much trust in.”

Infusing integrity into the supply chain needs to be part of the overall design scheme.

“The ideal way to solve this is at a fundamental level so all communication with a legitimate component can be encrypted and authenticated, and you can’t put a chip in between to intercept anything because it can’t read what’s been encrypted for another device,” Levine said. “But it has to be done early—at the foundry or in design. If you do it early enough you can follow it through the whole supply chain. “

That’s a tall order, and it’s generally well beyond the capabilities of even the largest IDMs.

“Supply chains for chips and trust devices are complex; they cover half the world before they end up in our hands,” according to Pim Tuyls, CEO of IntrinsicID. “Efforts like Siemens’ are definitely addressing the awareness and expectation of a certain level of security. But you’ll inevitably have devices in proximity to hardware from a company that isn’t in your group, or that you may not trust, that may have to make changes to your device. So if you find a chip in there, later you have to check all the parts that could have been added along the way and use special equipment to test what you find. It’s a lot of effort.”

The U.S. military also is facing something of a crisis in its Trusted Foundry program. The only foundries able to produce 7nm chips are TSMC and Samsung, and those are offshore. SMIC is working on that node as well, according to sources, and Intel is working on its 10nm process, which is roughly equivalent to 7nm for the foundries. But Intel reportedly is not interested in making chips for the government because volumes are too small.

GlobalFoundries, which canceled its 7nm development last year, is qualified for DoD production down to 28nm. To remain competitive, the military will need AI chips developed at advanced nodes, and so far there is no plan in place to make that happen.

“There are multiple solutions being considered,” said DARPA’s Leef. “The government is keenly aware and there is a lot of effort going into solving it. There are technology solutions, where you split the manufacturing and withhold the secret stuff, so basically you manufacture everything but the top layer. You also can do half in one place and half in another. This is a high-priority issue, though, and a lot of smart people are trying to solve it.”

Malicious hardware
Hardware Trojans are well known as a practical threat and are studied with increasing intensity as they become easier to make, thanks to design automation and reusable IP and the growing number of connected devices provides more targets.

“It’s not hard to believe that someone, maybe an employee, could be convinced to add a rogue element, a tiny little capacitor or something, to a board,” according Mentor’s Chen. “There was a bug we heard about that looked like a generic Ethernet jack, and it worked like one, but it had some additional cables. The socket itself is the Trojan and the relevant piece is inside the case, so it’s hard to see.”

Hardware Trojans can be a lot more dangerous than software because they’re able to bypass the root-of-trust provisions that keep firmware secure on a device, but they’re also harder to find, especially for companies trying to minimize cost by looking for the most likely bugs in their own designs, not scanning for bugs in places no one expected to look.

“It’s actually not that easy to find. As an industry we’re not in the habit of using x-rays to check for anomalies in every foreign board or look for a Trojan disguised as an Ethernet jack,” Chen said. “But it’s the same as anything else. It costs money to look and it’s like looking for a needle in a haystack—you can’t do that for every device in the world and I’m not sure how much you could catch. It’s actually easier to start from the software side and work backward.”

Traffic monitoring from inside devices can help, as well.

“Some of the techniques for heating up devices or adding glitches onto supplies can be picked up by the monitors, and from there you can determine if it’s been hacked,” said Stephen Crosher, CEO of Moortec. “These are signatures of malicious behavior. We’re providing that information to the software and the system above us. How that information is used is up to them. But step changes in temperatures or supplies would be noticed by the monitors. You’re talking tens of millicelsius being detected over a fairly rapid sampling time of tens of microseconds. If you’ve got something that’s working well at room temperature, and suddenly it hits minus 40, there’s either a fault or some malicious behavior.”

Unexpected behavior can be tracked through data movement, as well. “We have a customer using our [in-circuit monitoring] technology for malware detection, where they observe transactions and processes,” said Rupert Baines, CEO of UltraSoC. “They will label a process or software task as suspect right down to a very granular level. We’ve created a universal lock-step mechanism. We’re looking at two program streams, and if they diverge we sound an alarm. That’s happening on the fly, in real-time, in situ, while it’s running.”

And then there is good old-fashioned engineering to prevent problems.

“On-the-fly identification of suspicious transactions is important, but it is not a substitute for high-integrity ICs,” said Raik Brinkmann, president and CEO of OneSpin Solutions. “Hardware security pre-silicon verification is rapidly going up in the list of priorities of many engineering teams. One big challenge is that mainstream verification solutions focus on the intended use cases and what the hardware is supposed to do. Malicious agents, however, focus on misuse cases, leveraging bugs, vulnerabilities such as side channels, and possibly Trojans. IC integrity is about assuring not only functional correctness, but also the appropriate level of safety, security, and trust. With hardware sitting at the bottom of the stack, IC integrity forms the basis for the security of all the firmware and software layers on top.”

Cost reduction
All of this costs money, though, and even defense budgets are not unlimited. Uber-level questions about supply chain governance and reliability are being addressed by, among other efforts, the same DARPA programs trying to make chips cheap and easy enough to design in the US. The issue has to do partly with national security that could be damaged if U.S. sources were suddenly cut off from the fabs and facilities concentrated largely in China.

The long-term concerns of security during peacetime bring plenty of incentive to revamp a fragmented supply chain and bring quality and costs under control, however, according to Andreas Olofsson, Microsystems Technology Office program manager in charge of the Intelligent Design of Electronic Assets and POSH Open Source Hardware portions of the six-part program.

“If you look at the average $100 million to $200 million price tag, the cost is too much for non-commercial customers, but the cost is not the software or design tools, it’s the engineering,” Olofsson said. “We’re trying to find a way to reduce the engineering and cost through automation and re-use as close to zero as possible.”

That should bring design down into a range the DoD and other government agencies can manage and encourage the growth of domestic providers to balance out the chaotic fragmentation of the overseas market.

Other efforts aimed at this include DARPA’s two-year-old CHIPS program—aka, the Common Heterogeneous Integration and IP Reuse Strategies—which has focused on chiplets and standardized approaches to integrate those chiplets into a device. Whether this succeeds or not isn’t clear. The semiconductor industry is working on its own versions of this. One is being driven by IEEE’s International Roadmap for Devices and Systems, in conjunction with SEMI. There also is work by a group comprised of Netronome, Achronix, Kandou Bus, GlobalFoundries, NXP, Sarcina Technology and SiFive. And in Europe, Leti and Fraunhofer are working on a similar plan with others.

Marvell has been working on its own similar model internally, creating IP that can be put together off a menu based on a standard organic substrate. The company recently has expanded that to include third-party IP, as well.

“We’re no longer doing everything in-house,” said Yaniv Koppelman, networking CTO and senior director at Marvell. “Analog companies have mastered the complicated RF interface. There are two levels to that interface. One is the PHY, and on top of that is the MAC layer. We’re looking at CCIX as the next interconnect for that.”

The goal here is a LEGO-like approach, which can speed up chip development and reduce cost, but characterizing all of the different blocks sufficiently requires input from the entire supply chain as well as a grand plan for ensuring that the supply chain is secure enough and that no malicious hardware is added into the device.

Conclusion
Security is not a new problem, but the recognition that hardware is potentially as vulnerable as software is definitely a new twist. In the past, most side-channel attacks required grinding off the top of chips, examining where to insert probes with a scanning electron microscope. That’s still possible, but there are many more ways to grab critical data from chips than with brute-force attacks.

Coupled with rising cybersecurity concerns and widespread reports of hacking, both government and industry are taking this very seriously. It will still take years to fully implement, but for the first time momentum is building for secure hardware to run beneath secure software. And that is a monumental step in the right direction.

Related Stories
New Approaches To Security
Data analytics, traffic patterns and restrictive policies emerge as ways to ensure that systems are secure.
Finding Security Holes In Hardware
Emphasis on performance, backward compatibility and system complexity are creating vulnerabilities that are difficult to fix.
Using AI Data For Security
Pushing data processing to the edge has opened up new security risks, and lots of new opportunities.
Next Wave Of Security For IIoT
New technology, approaches will provide some protection, but gaps still remain.
Building Security Into RISC-V Systems
Emphasis shifting to firmware, system-level architectures, and collaboration between industry, academia and government.
Blockchain May Be Overkill For Most IIoT Security
Without an efficient blockchain template for IoT, other options are better.