Core Technology

Example: Fuzzing an image parser application

Practical Limitations of Fuzzing

Fuzzing is hard to integrate into existing development environments. The integration requires fuzzing and domain knowledge. Existing tools are usually from fuzzing experts for other fuzzing experts and therefore hard to use. These limitations have encouraged us to create a solution for developers and experts alike.

• Feedback-based fuzzing
  based on instrumentations for common -processor -architectures
• Support for the most important -architecture
  such as X86/64, ARM, AARCH64, MIPS or MIPS64
• No reliance on source code
  based on advanced binary translation and -instrumentation techniques
• Scalability
  of running tests in virtual environment
• Fuzzing via -different -interfaces
  like networks, files, and devices
  
• Virtualization and platform-agnostic
  due to the usage of QEMU (customizable -virtualization software)

• Simulation of widely used peripherals
  such as WIFI, Bluetooth, CANBus or Serial interfaces
• Low false-positive rate
  due to actual execution of the code rather than -static analysis based on patterns and data flows
  
• Structure-aware fuzzing
  to trigger deep states in programs that use custom data types such in network protocols. CI Fuzz comes with a set of predefined structures to handle several formats and protocols (JSON/XML/YAML/CSV/...).
• Stateful fuzzing
  Support for fuzzing stateful applications that require exchange of several messages of a given format to trigger deep states in the code.