Side-channel attacks are a class of exploit that infers sensitive information not through direct access to the data but from observing the nonfunctional characteristics, such as timing, power consumption, electromagnetic emissions, acoustic effects or cache behavior. Typical sensitive information that can be leaked includes cryptographic keys, keystrokes, webpage and video fingerprints, contents of a voice call, etc.

In this presentation, we will provide a brief overview of the world of side channels, by introducing you to side channels in networks, interconnects, caches, and other microarchitectural elements of a CPU.

We begin by exploring how side channels in networks can reveal the exact webpage or video the user is accessing, or what they are talking about on a VoIP call. Next, we introduce how a similar side channel in interconnects like PCIe and CXL can be exploited to identify which machine learning model is being used, or what a user is typing on their screen or into an SSH connection. We then examine how cache side-channel attacks leak cryptographic keys from the cache access patterns of a program. Finally, we discuss transient attacks, which extend cache attacks to read arbitrary addresses by leveraging the transient (e.g. speculative) execution.