Cache timing is a side-channel technique that extracts or transmits information by exploiting the difference in memory access time depending on whether data is present in the CPU cache, the fast memory where a processor keeps frequently used data. Also called a cache timing attack, it is one of the most studied classes of side channels and was first widely used to recover cryptographic keys.
Two processes sharing the same CPU and cache can exchange binary signals without any network or shared memory: one alters the cache state while the other measures access latency. This makes cache timing a covert channel that can bypass isolation in environments where hardware is shared, such as cloud virtual machines and containers. The Spectre and Meltdown vulnerabilities disclosed in 2018 also relied on cache timing to leak data.