mborgerson / mdec Goto Github PK

Decompilation as a Service. Explore multiple decompilers and compare their output with minimal effort. Upload binary, get decompilation.

Home Page: http://decompilerexplorer.org

License: Other

Dockerfile 15.82% Python 60.28% HTML 22.98% C 0.92%

decompilation ghidra angr ida binary-ninja reko retdec snowman decompiler jeb

mdec's Introduction

Update: mdec is now deprecated in favor of the community maintained decompiler-explorer project!

mdec

Explore multiple decompilers and compare their output with minimal effort. Upload binary, get decompilation.

Supported Decompilers

Notes:

Hex-Rays and Binary Ninja require license and binaries; other decompilers will be downloaded automatically.
JEB CE requires a license code. It's free. More info here.

Components

Each decompiler is a service that runs in its own container
A frontend web service proxies requests to backend service

Install

You'll need to add your proprietary packages in backend/*/private. Then just:

COMPOSE_DOCKER_CLI_BUILD=1 DOCKER_BUILDKIT=1 docker compose build

Tip: If you want to build only a few services, append the service names to the above command. For example, to build only the free/open source decompilers you can specify:

COMPOSE_DOCKER_CLI_BUILD=1 DOCKER_BUILDKIT=1 docker compose build \
  frontend \
  angr \
  ghidra \
  r2dec \
  reko \
  retdec \
  snowman

Run

docker compose up

Point your browser at http://127.0.0.1.

Tip: If you want to start only a few services, append the service names to the above command (e.g. frontend, angr, etc.)

API

You can also request decomp like:

curl -F '[email protected]' http://127.0.0.1/hexrays/decompile

mdec's People

Contributors

Stargazers

Watchers

mdec's Issues

Stand up publicly accessible service

Eventually this project will run publicly for people to use (like compiler explorer)
Some improvements need to be made first for scalability and platform abuse mitigation
Will continue to support local hosting

Support compiler explorer live source editing + compiling -> decompiling

Rename "IDA" to "Hex-Rays Decompiler"

Use official branding to avoid confusion

Show decompiler version

For ease of comparison, it would be great if the decompiler release/version was shown alongside the results. Then you could also compare different versions of the same decompiler and see the improvements over time.

IDA also has 2 versions: IDA version and Hex-Rays version, so you should show both.

Fix container privs

Support partial decomp (specify symbol)

I'd like to have a text box where symbol names or addresses can be entered, with autocomplete

Make API better

Return JSON data with more info if available
- Version
- Configured settings
- Log #12
- List of functions/symbols/types recovered
- Source
- Duration of time taken to decomp
Support decomp options
- #9
- #3
- Support other representations (Pythonish, Disassembly, LLIL, HLIL)
- Have universal options, and binary-specific options
- Return available options via query, or along with decomp

Make frontend smarter:

Have central upload spot (clients upload same binary for each decompiler)
Rate limit + captcha for public use
Request which decomps you want, then return a series of JSON objects containing response from each decompiler
Support multi part decomp updates
- Each service returns at different times, while servicing a request
Query backend status
Handle work queuing
Handle compiling for round trip decomp #2

Add decomp options

Thought: A little gear icon next to each decompiler. Click it to open a popup with options published by the service, such as:

Output representation
Simplifiers
Style preferences

Try to align decompilation output for ease of comparison

Make sure function ordering is similar in each decompilation (e.g. sort by address, etc)

Add CI

Add other decompilers

~~reko~~
~~snowman~~
rellic
~~retdec~~

Ghidra decomp is very slow

Ghidra decompilation is much slower than other tools. I'm sure startup costs can be brought down here.

Add option to enable/disable decompilers

Support decompiler bug reporting

Collect decompiler errors and notify vendors providing specimen and failure log

Add timeout

Add tests

Improve error handling

Show logs/stderr

Service multiple requests at once

The services currently block while it waits for a decomp to complete. We can do multiple at once however.

Add a License

Currently the sources in the repo are unlicensed and therefore it's not (legally) possible to fork or contribute changes. You should pick a license!

Offer pre-built container images

Some of the services take a long time to build. We can push these images from CI to ghcr for people to use

Refactor common server logic

Remote decompiler service

Some people may have computer-specific licenses of IDA / BN and are unable to run the decompilers on other machines. There should be a way to host a decompilation service on a remote machine and have the web server contact it for results.

Support Non-Linux IDA

Currently the IDA script creates a linux docker container and expects the linux version of IDA. Not everyone has a copy of the linux version of IDA, so it would be beneficial to support other platforms. May depend on #17 unless people are willing to run windows docker containers on windows hosts, and I'm not sure how macOS support would work.