Bindle

Dynamic testing, such as fuzzing, of software requires the development of an application-specific harness to provide inputs on all the input channels expected by the software. Today's state-of-the-art mutational fuzzing is also much more effective when it starts with sample, or seed, inputs that exercise some of the software's behavior.

Collecting good seed inputs requires an in-depth understanding of the application, while writing harnesses requires reverse-engineering and software-development expertise. Bindle eliminates these barriers, making dynamic testing much easier and faster to deploy. The tool's initial focus is on Linux applications.

Approach

Bindle collects seed inputs and generates harnesses automatically by learning from the tester simply running the software, as follows:

• Record
  • Observe execution - monitor the execution of the software, whether it is run manually or using a test script
  • Record inputs - observe the software's input channels, such as files, network sockets, and command line; collect observed input as candidate seed inputs
• Create
  • Generate a harness and seed inputs - suggest input channels for mutation and effective seed inputs to be reviewed by the tester
  • Aggregate inputs - combine seed inputs on all channels for mutation by fuzzer
• Run
  • Install harness - start the software with Bindle run-time interceding between the fuzzer and the software
  • Demultiplex inputs - separate the fuzzer's mutated input into component inputs on the software's input channels; supply data on demand.

Demo

The following video demonstrates the usage of Bindle to harness an application for fuzz testing.