CEO Outlook On Chip Industry

Part 3: The growing impact of security on design, and where the discontinuities and opportunities will be over the next five years.

popularity

Semiconductor Engineering sat down with Wally Rhines, president and CEO of Mentor, a Siemens Business; Simon Segars, CEO of Arm; Grant Pierce, CEO of Sonics; and Dean Drako, CEO of IC Manage. What follows are excerpts of that conversation. To view part one, click here. Part two is here.


L-R: Dean Drako, Grant Pierce, Wally Rhines, Simon Segars. Photo: Paul Cohen/ESD Alliance

SE: Security is a growing issue, and it’s a problem that needs to be addressed from many angles. How do we figure out where there are vulnerabilities and eliminate them?

Rhines: There is every possibility and every likelihood there will be embedded Trojans and other things that get into chip designs. So far nobody has been hurt and there hasn’be been physical damage.

Drako: That’s because 99.99% of all security issues have been in software.

Rhines: There are tools available to design in higher levels of security for the silicon. People have enough to worry about with the software and all the different layers before they get down to the silicon, so they’re not really motivated yet to spend a lot of time building security into the silicon itself.

SE: Haven’t Meltdown and Spectre changed that?

Drako: They are vulnerabilities. They weren’t really attacks.

Segars: They do bring to light a key issue, though, which is that you can be happily doing something for decades thinking it’s safe, and then somebody finds an attack mechanism that no one else thought of.

Drako: I deal with the security industry, where more than 90% of the security cameras are manufactured in China. The chips in those cameras used to be made in a lots of different places. They’ve since migrated to China, and a lot of the government customers I engage with are unwilling in any way, shape, or form to deploy those cameras. They have a huge problem sourcing cameras that are not based on those chips. There is a lot of concern about the Trojans in chips and Trojans near the chips. It’s the first situation I’ve encountered where the customer is honestly concerned about this.

Rhines: We have lots of technology that could be implemented. That includes PUFs in chips and camouflage. There are design tools for this, too. But it takes effort. And if you’re not designing chips for the NSA, you’re just not that motivated to spend the money and time today. But somewhere down the road it’s going to become a liability, and then people are going to get very interested.

Segars: Right now it depends on the end market for whether people actually care about security. And generally, if there’s a person or a relatively small number of people who can quantify the loss and who have full control over their supply chain, they can dictate how things get built. The movie industry really cares about losing its content. If you’re developing a CD player, they will ask whether you have an X, Y or Z chip in there. If you don’t, then you can’t have their software. The banking industry is similar. There’s a consortia of banking institutions who say, ‘We’re all at risk here if credit cards get tampered with, so lay down some standards for how secure tiles get built into banking applications.’ For those examples, it’s relatively easy. But for IoT at large, I spend a lot of time worry about whether I’m going to get paid more for having a secure chip. A lot of people don’t, so they cut corners. But that is going to have to change if consumers are going to feel safe deploying something in their homes. If you buy something at a store and hook it up to your home network, you have no clue whether the person who built that spent zero minutes or thousands of hours worrying about security. You have no way to tell that.

SE: And that’s one of the big issues with the IoT. It’s not necessarily your device. It’s what you connect with. That was one of the issues with the Mirai attack, right?

Segars: Yes, particularly the software that’s put on top. You can have all the security features you like, but if someone builds a system where the user name is ‘admin’ and the password is ‘password,’ then you’re just asking for trouble. There’s nothing a chip designer can do to stop that. Most people don’t think they’re at risk of getting hacked, and most people don’t think they’re interesting to a hacker. But your house may have been used as part of that Mirai botnet attack, and it’s possibly still being used.

Pierce: When we have Google and Facebook collecting such a deluge of data about us, and we interact with it, that suddenly creates a public awareness about whether the device I’m using really needs to capture all of that data and share it with the cloud. These big companies can track cell phone calls, too. They know everything. You may see devices where you pay extra for a secure phone. It’s not necessarily that you’ve been attacked. It’s the idea that you don’t have any privacy.

Drako: That’s parly a generational thing. Younger people accept that’s the way the world is. They’ve grown up with this. But we are heading toward more encryption of data. The social media world is beyond the Wild West. We’re going to learn from it and rein that it. There’s a whole societal learning that’s taking place.

Pierce: In the old days, we used to fire up a design and it was all about performance. Then it became performance and area, and more recently it extended to power. In the future it will extend to security and form factor. You’re not just creating an embedded chip. It will have to play across all of these different planes of the design decision.

SE: Where do you see the industry moving over the next five years, and where do you see discontinuities?

Segars: The challenge of understanding every component in the system, and how a design is done and when to sign off, will have to shift. With safety-critical and security-critical, the supply chain will have to become a lot tighter. We will all have to prove more about what we do and show our work, not just the end result. There will be a lot more applications that require that for EDA and IP.

Rhines: About 80% of the growth of the EDA industry is for new capabilities, while only 20% of the growth is for usage of traditional capabilities. So as long as we can count on this new complexity coming along and new problems, such as security and power and thermal analysis, that will create growth. But the other thing I would point out is that the application of EDA techniques, simulation, and verification to IC design is 40 years ahead of the application of those same techniques to system design. For the EDA industry and where it goes, solving the problems of how you automate the electronic design of cars, planes, trains and machinery is going to grow at twice the rate of the growth of the intrinsic IC design business. It’s an enormous opportunity.

Pierce: The customers coming into to the market today are increasingly focused at a system level. A lot of times these are software guys or systems people trying to create these end devices. They can’t consume what we provide at the level we provide it. They definitely are going to need a way of abstracting up because they won’t have the time to design, optimize and implement every transistor. They won’t have the time to design and optimize every IP core. So we’re going to have to bring these things together in a form they can consume, maybe as subsystem platforms, with software already available, and perhaps delivered together. That’s a big change.

Drako: I don’t think there will be any discontinuities in the EDA industry.

Rhines: You don’t think the cloud is a discontinuity?

Drako: No. We’ve been going up and to the right for 40 years. It’s not like the stock market. But I do think there are two major influencers that will drive a whole new set of tools. One is that there is a slew of companies doing neural network chips, and those are delivering things people use every day in the form of Alexa or image classifiers. There’s going to be a lot of money put into optimizing those algorithms in chip design, and that’s going to require a whole new set of tools. We have automated tools for designing hardware layouts of memory and memory compilers. We’re going to have neural network compilers and things like that. And the cloud is going to shift the industry. But it’s going to be gradual. Things happen a lot more slowly and gradually than people would ever have imagined.



2 comments

Dave Noble says:

The only way to truly implement security in an IC is with hardware root of trust to authenticate the device.

It has to encompass a mechanism like Physical Unclonable Devices whereby the physics of the SRAM are used to create a unique ‘fingerprint’. This means that no key exists so nothing is hidden on the device. Side channel attack countermeasures and Error Correction also need to be implemented to make the PUF resilient to ‘back door’ attacks.

Tom says:

Regarding the Mirai attack, the security problem lie in the structure of the security camera industry, and the consumer electronics industry at large. Suppliers like Xiongmai are paid to build hardware for a set price, which usually gets rebranded under another company’s name. The rebranding company does the marketing to consumers, bears any responsibility for faulty products, and makes the profits on them. Since Xiongmai makes no money once products leave the warehouse, it’s not incentivized to care that much about how well they’re made. And that’s a source of the “admin” and “password” defaults. This is why chip-identity driven authentication has a lot of potential.

Leave a Reply