Microarchitectural attacks such as Spectre and Meltdown have shown that the optimizations built into modern processors—caching, speculative execution, and the like—can leak information through software-visible side effects. These attacks undermine security because the hardware's behavior is not fully specified to software, leaving gaps that attackers can exploit.

The new paper, "Foundations of Hardware-Software Leakage Contracts," argues that to use modern hardware securely, we need explicit agreements about what information may leak. The authors propose formal "leakage contracts" between hardware and software, analogous to the instruction-set architecture (ISA) contracts that define functional behavior. Such contracts would specify the permitted side channels and the conditions under which they can be observed.

By making leakage properties explicit, these contracts could give software developers a solid basis for reasoning about security, and hardware designers a clear target for verification. The work is at an early stage, but it points toward a principled way to address a class of vulnerabilities that has proven resistant to ad-hoc fixes.