HiFiNet: Hierarchical Fault Identification in Wireless Sensor Networks via Edge‑Based Classification and Graph Aggregation

Dependencies

• Python
• Pip
• PDM

Installation

Clone the repo:

git clone https://github.com/nghiant03/HiFiNet

Navigate to the repo directory. Then install the repo as a package with:

cd HiFiNet
pdm install

Optional: Install linting and lsp depedencies for using type check. HiFiNet repo is strongly linted and type hinted:

pdm install -G:all

Usage

HiFiNet experiments can be run using the intergrated CLI.

hifinet --help

Inject

There are default adaptor for the 2 datasets used in the paper. Therefore the dataset can be downloaded and placed inside the directories inside data/. The injecting command for each will then be simply:

hifinet inject intel
hifinet inject merra2

The injection functionality can be used with any dataset .csv file. The data need to be in the following schema:

• node_id column: Unique ID for each node of the WSN
• target column: The target feature needed to be injected
• feature/feature_* columns: The additional features presented in the data

hifinet inject path/to/dataset

To customize fault parameters, supply a JSON file via --fault-config:

hifinet inject path/to/dataset --fault-config path/to/config.json

config.json should map fault names (for example, drift or spike) to their parameter overrides:

{
  "drift": {"sigma": 0.02},
  "spike": {"bias_range": [2.0, 4.0]}
}

Inline JSON strings are supported for quick tweaks:

hifinet inject path/to/dataset --fault-config '{"drift": {"sigma": 0.02}}'

See the available option to configure the injection functionality:

hifinet inject --help