DRILL

Log-based Anomaly Detection for Unseen Logs via Source Code Feature Extraction

Description of File

Drill.ipynb: this file is the evaluation part of the paper IPDPS'23 Log-based Anomaly Detection for Unseen Logs via Source Code Feature Extraction

features.pkl: this file is the sentiment and context features of HDFS log statements.

data: this directory contains the log sessions of HDFS, the corresponding log templates of log indices are described in Drill.ipynb.

figures.ipynb: this file displays the figures in the paper.

sentilog: this directory contains the code of paper HotStorage'21 SentiLog: Anomaly Detecting on Parallel File Systems via Log-based Sentiment Analysis., which shows how we extract the sentiment features in Drill.

Dependencies

pytorch: pip3 install torch

d2l: pip3 install d2l

sklearn: pip3 install scikit-learn

jupyter notebook: pip3 install notebook

How to Run

jupyter notebook Drill.ipynb