ShowTime: Amplifying Arbitrary CPU Timing Side Channels.

Microarchitectural attacks typically rely on precise timing sources to uncover short-lived secret-dependent activity in the processor. In response, many browsers and even CPU vendors restrict access to fine-grained timers. While some attacks are still possible, several state-of-the-art microarchitectural attack vectors are actively hindered or even eliminated by these restrictions. This paper proposes ShowTime, a general framework to expose arbitrary microarchitectural timing channels to coarse-grained timers. ShowTime consists of Convert routines, transforming microarchitectural leakage from one type to another, and Amplify routines, inflating the timing difference of a single microarchitectural event to make it distinguishable with crude sources of time. We contribute several Convert and Amplify routines and show how to combine them into powerful attack primitives. We demonstrate how a single cache event can be amplified so that even the human eye can classify it with 98% accuracy and how stateless time differences as minuscule as 20 ns can be captured, converted, and amplified in a single observation. Additionally, we generate cache eviction sets, both in real-world restricted browser environments and natively using timers with precisions ranging from microseconds to seconds. Our findings imply that timer restrictions alone, even when ruthlessly implemented beyond practical limits, provide insufficient protection against CPU timing attacks.