internal/fuzz: deduplicate interesting inputs #48303
Labels
fuzz
Issues related to native fuzzing support
NeedsInvestigation
Someone must examine and confirm this is a valid issue and not a duplicate of an existing one.
Milestone
When the fuzzing engine starts, it runs inputs in the seed corpus and cached inputs to gather baseline coverage.
If two inputs provide the same coverage, we should prioritize one and ignore or deprioritize the other, perhaps deleting it if it was cached.
The AFL whitepaper explains how they deduplicate inputs based on coverage. We might use a similar algorithm. We could also use size and duration as metrics: smaller, faster inputs are better.
The text was updated successfully, but these errors were encountered: