goMidas

Go implementation of C++ code by Siddharth Bhatia

Installation

You can install and use the package by cloning this repository in your project folder:

git clone https://github.com/ritesh99rakesh/goMidas.git

Table of Contents

• Features
• Use Cases
• Installation
• Demo
• Datasets
• MIDAS in other Languages
• Online Articles
• Citation

Features

• Finds Anomalies in Dynamic/Time-Evolving Graphs
• Detects Microcluster Anomalies (suddenly arriving groups of suspiciously similar edges e.g. DoS attack)
• Theoretical Guarantees on False Positive Probability
• Constant Memory (independent of graph size)
• Constant Update Time (real-time anomaly detection to minimize harm)
• Up to 48% more accurate and 644 times faster than the state of the art approaches

For more details, please read the paper - MIDAS: Microcluster-Based Detector of Anomalies in Edge Streams. Siddharth Bhatia, Bryan Hooi, Minji Yoon, Kijung Shin, Christos Faloutsos. AAAI 2020.

Use Cases

1. Intrusion Detection
2. Fake Ratings
3. Financial Fraud

Usage

There are two ways to use goMidas:

1. Use the Midas and MidasR functions in the package in this repository:

   Example:

   package main
   
   import (
       "fmt"
       goMidas "github.com/ritesh99rakesh/goMidas"
   )
   
   func main() {
       src := []int{2, 2, 3, 3, 5, 5, 7, 11, 1, 2}
       dst := []int{3, 3, 4, 4, 9, 9, 73, 74, 75, 76}
       times := []int{1, 1, 2, 2 ,2, 2, 2, 2, 2, 2}
       numRows := 4
       numBuckets := 769
   
       // get anomaly scores
       anomalyScore := goMidas.Midas(src, dst, times, numRows, numBuckets)
       fmt.Println(anomalyScore)
   }

2. Use the main.go provided in example folder in this repository to run MIDAS algorithm from command line.

   The file main.go just requires you to specify the data.csv file containing containing src, dst and timestamps.

   You just have to run:

   go run main.go -input <input-file> -<other-optional-arguments>

   Complete details for the available arguments:

   -alpha float
       Alpha: Temporal Decay Factor. Default is 0.6 (default 0.6)
   -buckets int
       Number of buckets. Default is 769 (default 769)
   -input string
       Input File. (Required)
   -norelations
       To run Midas instead of Midas-R.
   -output string
       Output File. Default is scores.txt (default "scores.txt")
   -rows int
       Number of rows/hash functions. Default is 2 (default 2)
   -undirected
       If graph is undirected.

   For more details and to know how to find AUC of the anomaly scores, refer example folder in this repository.

Datasets

1. DARPA: Original Format, MIDAS format
2. TwitterWorldCup2014
3. TwitterSecurity

MIDAS in other Languages

1. C++ by Siddharth Bhatia
2. Python by Ritesh Kumar
3. Rust by Scott Steele
4. Ruby by Andrew Kane

Online Articles

1. KDnuggets: Introducing MIDAS: A New Baseline for Anomaly Detection in Graphs
2. Towards Data Science: Controlling Fake News using Graphs and Statistics
3. Towards Data Science: Anomaly detection in dynamic graphs using MIDAS
4. Towards AI: Anomaly Detection with MIDAS

Citation

If you use this code for your research, please consider citing our paper.

@article{bhatia2019midas,
  title={MIDAS: Microcluster-Based Detector of Anomalies in Edge Streams},
  author={Bhatia, Siddharth and Hooi, Bryan and Yoon, Minji and Shin, Kijung and Faloutsos, Christos},
  journal={arXiv preprint arXiv:1911.04464},
  year={2019}
}

Issues

If you find any problem with the code, please raise an issue.