Hardware-assisted Adaptive Cross-Layer Security for Computing Systems

Today's computing systems are facing an unprecedented security threat posed by recent attacks that use software to exploit hardware vulnerabilities, as shown by attacks like Spectre, Meltdown, Foreshadow, and follow-ups - affecting a wide range of computing platforms and manufacturers, including Intel, AMD, and ARM.

These cross-layer attacks reach far beyond exploiting microarchitectural vulnerabilities and allow unprivileged software to exploit a variety of hardware design and implementation flaws, as we demonstrated in the world's largest System-on-Chip (SoC) security competition that we have been conducting with Intel since 2018.

This adversarial paradigm shift sidesteps decades of security research that assumed a layered architecture where hardware is flawless and trustworthy.

Existing solutions, such as software patching or specific hardware changes are ad-hoc, expensive, or only mitigate specific known attacks.

Particularly, patching hardware after fabrication is very limited or impossible.This proposal, HYDRANOS, envisions hardware-assisted adaptive security, a radically different approach to enable flexible security for future computing systems.

We aim to design, prototype, and evaluate dedicated configurable hardware inside the SoC design to enable post-fabrication reconfiguration of key security-relevant hardware primitives to mitigate new attack vectors.

Moreover, we provide an evaluation framework that includes novel hardware fuzzing techniques to significantly improve existing hardware-vulnerability detection methods at design time.HYDRANOS is a game changer for trustworthy computing, allows to fundamentally and flexibly tackle todays and future cross-layer attacks on security-critical systems, and provides novel research to pave the way towards future-proof security.

We will showcase our results on open-source hardware widely supported by academia and industry and provide it to the research community, allowing open verification by third-parties.