The unleashed Cyber-Physical Systems (CPS) and Internet of Things (IoT) revolution
creates exceptional opportunities in many domains such as healthcare, smart power
grid, automotive, industrial robots, and transportation systems. Systems-on-Chip
(SoCs) present appealing platforms that enable this revolution due to their performance,
power, and area (PPA) advantages in addition to their heterogeneity, leading to Cyber-
Physical Systems-on-Chip (CPSoCs). However, CPSoCs are not possible without
addressing the one big concern associated with the prevalence of CPS: safety. Safe
CPSoCs must be predictable, secure and verifiable. 1) Predictability ensures the satisfaction
of timing requirements. Unpredictable CPSoCs can fail to meet timing deadlines,
which leads to severe consequences such as life losses (think in the air bag in your car).
However, current commodity SoCs are not predictable. They adopt numerous architectural
optimizations that offer high-performance at the cost of huge variability in timing.
2) Security is one of the biggest challenges facing architects of CPS. Unlike traditional computing systems, CPS manage sensitive tasks; therefore, any security breach could
lead to catastrophic consequences. These consequences range from revealing personal
information (e.g., from wearable devices) to a global threat (e.g., compromising a
nuclear plant). Consequently, ensuring the security of CPSoCs is a first-class mission.
3) Verifiability, reliability and correct functionality in CPSoCs are of a vital importance.
Think for example in self-driving cars, self-adaptive home appliances, or smart power
grids. Similar to untimeliness, malfunctioning in these systems can also lead to life
losses. Accordingly, the rigorous verification of each system component is unavoidable
process for these markets to successfully scale.
In this talk, I present several novel solutions to provide predictable, secure, and verified
CPSoCs. Finally, I highlight research directions towards enabling the widespread of
CPSoCs and open problems yet to be addressed.
