Skip to content
Hadoop and Storm based outlier analysis implementations for cyber security and fraud detection
Java Shell Ruby Scala Python
Latest commit 9035d8b Mar 27, 2016 Pranab Ghosh config param key prefix
Failed to load latest commit information.
resource config param key prefix Mar 26, 2016
spark spark port intial commit Oct 6, 2014
.gitignore spark artifact Oct 6, 2014 refactoring of meta data classes Jul 13, 2015 initial commit Jan 3, 2012
pom.xml use jackson asl license Oct 11, 2014


Beymani consists of set of Hadoop and Storm based tools for outlier and anamoly detection, which can be used for fraud detection, intrusion detection. All the implementations will be ported to Spark.


  • Simple to use
  • Input output in CSV format
  • Metadata defined in simple JSON file
  • Extremely configurable with tons of configuration knobs


The following blogs of mine are good source of details of beymani


  • Multi variate instance distribution model
  • Multi variate sequence or multi gram distribution model
  • Average instance Distance
  • Relative instance Density
  • Markov chain with sequence data
  • Instance clustering
  • Sequence clustering

Getting started

Project's resource directory has various tutorial documents for the use cases described in the blogs.


For Hadoop 1

  • mvn clean install

For Hadoop 2 (non yarn)

  • git checkout nuovo
  • mvn clean install

For Hadoop 2 (yarn)

  • git checkout nuovo
  • mvn clean install -P yarn


Please feel free to email me at


Contributors are welcome. Please email me at

Something went wrong with that request. Please try again.