0x00A0glossary
Glossary.
The terms this manual uses, with a short definition and the chapter where each one is explained.
0x00A8terms
- AFL++
- A coverage-guided fuzzer. Every runnable target in Parsarium has an AFL++ overlay. Ch. 5
- Adoption rank
- A package's position by one popularity metric, such as downloads or stars, among the packages in the same cell. Ch. 4
- Bridge
- The component that records parser source coverage in Go, Java, JavaScript, Python, or PHP and writes it into the AFL++ map. Ch. 5
- Cell
- One format in one language, such as PNG in Rust. Parsarium has 112 cells. Ch. 1
- Constraint
- A rule that relates parts of an input, such as a length field matching the data size or a checksum matching the bytes it covers. Ch. 2
- Control
- A small valid input with a known expected output, used to verify that a target works. Ch. 3
- Corpus
- A set of inputs: seeds given to a fuzzer, the queue it builds, or a generated set replayed through targets. Ch. 7
- Differential testing
- Sending the same input to several implementations and comparing their answers. Ch. 6
- Edge
- A transition between two code locations. AFL++ counts edges in its map to measure coverage. Ch. 5
- Parser family
- The code that actually reads the format behind a package. Many packages can share one family. Ch. 4
- Feedback kind
- How a target reports coverage to AFL++: compiled edges, or a bridge for a given language. Ch. 5
- Grammar
- A description of the parts of an input and their order and nesting. Ch. 2
- Harness
- The small program that reads one input, calls the parser's documented API, and prints what it observed. Each target's harness runs in a pinned Docker image. Ch. 3
- Lead
- A candidate that has not passed every check yet: a search hit in the catalog, or a saved crash or hang in a campaign. Ch. 8
- Input API
- The documented function a package provides to parse the format, such as
png_image_begin_read_from_file. Ch. 3 - Package path
- One package or standard API together with its input API. The unit counted in the catalog. Ch. 1
- Receipt
- A JSON record of what happened: a verification receipt for a target, a run receipt for one input, a summary for a batch. Ch. 3
- Seed
- An input a fuzzing campaign starts from. By default, the target's control. Ch. 7
- Selection ID
- An identifier the bench derives from a sorted list of targets. With the Parsarium commit, it pins a target set. Ch. 9
- Request smuggling
- An attack that uses two HTTP parsers disagreeing about where a request ends. Ch. 6
- Standard API
- A parser shipped with a language's standard library or runtime, such as Go's
image/png. Verified like packages, listed separately. Ch. 4 - Target
- A verified package path with a runnable Docker harness. Ch. 3