UAC-AD

Source code for the paper "UAC-AD: Unsupervised Adversarial Contrastive Learning for Anomaly Detection on Multi-source Data"

Environment

We support python3.x $\geq$ 3.7. The environment can be built by: $ pip install -r requirements.txt

Result records

The result records are in the result21 directory.

Reproducing UAD by running:

cd codes && python run.py

The overview of UAC-AD

Main Result

Experiment data types

Raw data for Dataset A: https://doi.org/10.5281/zenodo.7609780. The metric types for Dataset A include CPU status, memory status, IO status, and network status. The log type for Dataset A is Spark runtime logs.

Raw data for Dataset B: https://github.com/CloudWise-OpenSource/GAIA-DataSet/tree/main/MicroSS. The Dataset B is mainly comes from a scenario in the business simulation system, MicroSS, owned by Cloudwise. It comes from a scenario of logging-in with QR Code.

The data type for Dataset C is restricted due to confidentiality requirements and is not disclosed at this time.

Tree

.
├── README.md
├── codes
│   ├── common
│   │   ├── __init__.py
│   │   ├── data_loads.py
│   │   ├── data_processing.py
│   │   ├── data_processing_utils.py
│   │   ├── semantics.py
│   │   └── utils.py
│   ├── data_analysis.py
│   ├── gpu0.sh
│   ├── gpu1.sh
│   ├── models
│   │   ├── basev3.py
│   │   ├── fuse_v3.py
│   │   ├── kpi_model_v3.py
│   │   ├── log_model_v3.py
│   │   └── utils.py
│   └── run.py
├── data
│   └── chunk_10
│       ├── test.pkl
│       ├── train.pkl
│       ├── unlabel.pkl
│       └── unsupervised.pkl
├── requirements.txt
└── result21
    ├── main_result.png
    ├── overview.png
    └── test.txt