abstraktor

Abstraktor is a tool designed to assist in the validation of distributed protocol implementations by automatically generating enabling-preserving abstractions (EPAs). Unlike existing tools that only detect runtime errors, Abstraktor aims to uncover issues in the business logic of implementations.

The solution is based on four main components:

1. Instrumenter: Processes the source code of the target application to generate an executable that exposes its internal state.
2. Mallory Extension: A fork of the Mallory tool that performs efficient fuzzing while also extracting the internal state of the instrumented executable.
3. EPA Generator: Builds and exports EPAs using the state data collected during fuzzing.
4. CLI Tool: Coordinates and orchestrates the interaction between all components to produce the final abstractions.

Installation

To install Abstraktor, you need to have Rust and Cargo installed on your system. If you haven't installed them yet, you can follow the instructions on the official Rust website.

Once you have Rust and Cargo installed, you can clone the repository and build the project:

cargo build

Running tests

Run tests using:

cargo test

Workflow

sequenceDiagram
    participant Dev as Developer
    participant CLI as Abstraktor CLI
    participant LLVM as LLVM Instrumenter
    participant Mallory as Mallory Engine
    participant EPAGen as EPA Generator

    Dev->>Dev: Annotate source code
    Dev->>CLI: Run prepare-binary
    CLI->>LLVM: Instrument code using annotations
    LLVM-->>CLI: Instrumented binary ready

    Dev->>CLI: Run run-mallory
    CLI->>Mallory: Execute instrumented binary
    Mallory-->>CLI: Collect runtime events

    Dev->>CLI: Run generate-epas
    CLI->>EPAGen: Generate EPAs from collected events
    EPAGen-->>Dev: Output EPAs

Loading

Description

1. Annotate Source Code: Manually add comments to relevant parts of your application logic to mark key points for analysis.
2. prepare-binary: Instruments the source code using LLVM based on the provided annotations.
3. run-mallory: Executes the instrumented binary under Mallory to simulate network conditions and collect runtime events.
4. generate-epas: Processes the collected data to automatically construct enabling-preserving abstractions (EPAs).