readme.md

Integration of Malware Dataset into Fedstellar

To integrate the malware dataset into Fedstellar, follow these steps:

• Copy the malwares folder into the directory fedstellar/learning/pytorch/malwares.

• Modify the file fedstellar/node_start.py as follows:

  • Add the necessary import statements.
  • Include the malware option.

  from fedstellar.learning.pytorch.malwares.malwares import MalwaresDataset
  from fedstellar.learning.pytorch.malwares.models.mlp import MalwaresModelMLP
  
  ...
  
      elif dataset == "MALWARE":
          dataset = MalwaresDataset(
              sub_id=idx, number_sub=n_nodes, iid=iid
          )
          if model_name == "MLP":
              model = MalwaresModelMLP(
                  input_size=dataset.train_set.__getitem__(0)[0].shape[0]
              )
          else:
              raise ValueError(f"Model {model} not supported")

• Organize the data files from training/data/cleaned into the following structure:

  fedstellar/data
  ├── user
  │   ├── 0.csv
  │   ├── 1.csv
  │   ├── 2.csv
  │   ├── 3.csv
  │   ├── 4.csv
  │   ├── 5.csv
  │   ├── 6.csv
  │   └── 7.csv
  └── merged_df.csv

• Add the option <option>MALWARE</option> to the frontend HTML file located at fedstellar/frontend/templates/deployment.html:

              <select class="form-control" id="datasetSelect" name="dataset" style="display: inline; width: 50%">
                  <option selected>MNIST</option>
                  <option>FashionMNIST</option>
                  <option>CIFAR10</option>
                  <option>SYSCALL</option>
                  <option>SYSCALL</option>
                  <option>MALWARE</option>
              </select>

• Optional steps:

  • To test the dataset and run the model locally without Docker, perform the following:
    • Add the test_malwarefile to fedstellar/test_malware.py
    • Edit the path to a valid configuration file and run the test.
  • To generate the documentation, place the start_doc.sh script in the root directory.