Security, machine learning, and variety at DAC

By Chris Edwards |  No Comments  |  Posted: May 13, 2019
Topics/Categories: Blog - EDA, Embedded  |  Tags: , , ,  | Organizations: , ,

Security and machine learning are two topics that take center stage at the Design Automation Conference (DAC) this year, says the conference’s general chair Rob Aitken.

Aitken, who is an Arm fellow and director of technology at the company, says DAC has become a key conference for security innovations in electronic design. “We’ve got to the point where it’s quite self-sustaining as a topic. Eight years ago, we made a push to get more security material, which required actively going out and recruiting papers. Now the community is treating DAC as a useful venue for hardware security issues.”

As well as research into detecting hardware Trojans and defeating side-channel attacks, obfuscation is one of the themes in security this year, Aitken notes. “How you can make your circuitry be less vulnerable to reverse-engineering attacks. Another one is where your circuit doesn’t do anything until the circuit keys are loaded into it,” Aitken says. “It’s a big area of concern, particularly for the defense-industry people, when we don’t know who will be manufacturing the chips. Getting obfuscation that actually works [to hide IP] is a very challenging problem and it’s keeping people busy.”

The other key theme is machine learning and its application in EDA, Aitken says. He says the conference has seen strong organic growth “just because it’s part and parcel of how machine learning has taken over a lot of computing”. But for DAC, there is now a lot of activity in finding methods to enhance existing design techniques.

A surprisingly common usage model is adapting the deep neural network (DNN) structures originally developed for image and speech processing to EDA problems, and not necessarily for things that process image-like data, such as layout analysis. Aitken points to an example at his own company where DNNs have been applied to analyzing test-vector effectiveness in verification.

“A second model is using machine learning to create fitness functions or additional heuristics to be used in a standard EDA tool, to see whether, for example, we can make a better router,” Aitken says. “The third is layered on top of existing tools. If you take most EDA tool flows they have a fair number of tweakable knobs. By applying machine learning, you can have a tool that learns the best settings for the designs.”

Although this year’s DAC has a strong focus on topics like machine learning and security, Aitken is keen to stress the variety of content. The keynotes this year extend further beyond the day-to-day experience of design to encompass bigger themes, such as the ethics of AI, which is covered by Arm’s Carolyn Herzog in a Tuesday, June 4 SKYTalk.

“A deliberate decision was to use the keynotes as a means of broadening people’s horizons and bringing different perspectives in,” Aitken says.

Also on Tuesday is a keynote by musician Thomas Dolby. “He’s always been interested in the intersection of music and technology. He’s been dabbling in machine learning as well.

Alongside the keynotes and research papers, DAC continues its highlighting of real-world design issues through the Designers Track. “It’s been a conscious effort over a number of years to make more content at DAC reflect the work of practicing design engineers. That content is pretty strong this year,” Aitken says, pointing out the additional value of networking around the talks. “It remains the number-one reason why conferences exist.”

Comments are closed.


Synopsys Cadence Design Systems Siemens EDA
View All Sponsors