Requirements

Building

Create a venv e.g python3 -m venv maude and activate it.
Clone the repo and submodules: git clone https://github.com/allisterb/maude --recurse-submodules
Run install.cmd on Windows or install.sh on *nix/macOS. On systems with a small amount of memory like a VPS you may get an error like install.sh: line 5: 382383 Killed pip install ext/nsfw_model in which case you should use install_lowmem.sh.
Download models.zip or models.tar.gz from the latest release and unzip/untar in the root repo directory so you have a models directory alongside doc and test et.al. Note this step is only required when installing directly from the git repo; the release archives already include the models folder.
To use Microsoft PhotoDNA hashing you must put the PhotoDNAx64.dll (Windows) or PhotoDNAx64.so (Mac) where the maude Python interpreter can find it: e.g. on Windows in the Scripts folder of your maude Python venv. See here for more info.
On Windows the libclamav* DLLs are bundled with maude however on other platforms you'll have to install it yourself using your package manager e.g. sudo apt-get install libclamav-dev or other method. See https://github.com/Cisco-Talos/clamav/blob/main/INSTALL.md for more info.
Run maude or maude.sh from the root repo folder.

Use the start_ipfs scripts to start an IPFS instance e.g ./start_ipfs.sh ipfs.log ../../go-ipfs/ipfs or set the appropriate environment variables and flags yourself based on what the script says. IPFS needs to run with the --enable-pubsub-experiment flag and JSON log output. See the script for full details.
If this is the first time you are running maude you will need to generate a keypair for your instance: maude[.sh] crypto gen.
To start the maude server in monitor mode say maude[.sh] server monitor. This will start monitoring a local IPFS node using the node's logs and will detect when a CID is pinned locally. maude tries to detect the type of file being pinned like a binary executable or archive or a document like a pdf or a media file like an image or video. It then runs different classifiers and detectors on the file like computer vision models that can detect NSFW images and videos, and malware detectors like ClamAV and YARA rulesets on binary executables and archives. maude publishes the raw classification data to an IPFS pubsub topic.
You can subscribe to the topic e.g ipfs pubsub sub maude to see the data maude publishes on pinned files.
There are other maude commands for running the ML models and scanners on local files e.g. maude image classify --nudenet /path/to/a/image See maude --help for the full list of commands.