Feedback-guided input mutation

Feedback-guided input mutation is a fuzzing technique in which test inputs are mutated using information derived from instrumented execution of the target binary. The technique extracts feedback during execution and uses that feedback to guide how inputs are mutated, with the goal of generating test cases that exercise new program behavior.

In emulation-based fuzzing, where binaries are executed without source code and often on instruction set architectures (ISAs) different from the host, the integration of feedback-guided input mutation depends on the availability of instrumentation inside the emulator. Modern emulation-based fuzzers have historically evolved by re-purposing general-purpose emulators, which makes it difficult to add instrumentation in a uniform, architecture-agnostic way. As a result, support for feedback-guided input mutation is often added in an ad-hoc manner tied to specific ISAs, limiting the technique's reach on diverse firmware targets.