The effects of synthetically augmented training data on parameter tuning for anomaly detection algorithms

Leonard Lightfoot; Ellen Laubie; Joseph Natarian

doi:10.1117/12.883526

23 May 2011 The effects of synthetically augmented training data on parameter tuning for anomaly detection algorithms

Leonard Lightfoot, Ellen Laubie, Joseph Natarian

Proceedings Volume 8062, Defense Transformation and Net-Centric Systems 2011; 80620M (2011) https://doi.org/10.1117/12.883526
Event: SPIE Defense, Security, and Sensing, 2011, Orlando, Florida, United States

Abstract

While many years of research have been dedicated to anomaly detection algorithms and their applications, little research has been devoted to the act of tuning parameters to perfect the performance of these algorithms. This paper investigates three anomaly detection algorithms (Local Outlier Factor, wavelet decomposition, and a simple sliding threshold) and the effect of synthetically augmented training data on the resulting false positive and false negative rates. Four datasets were developed by injecting varying quantities of synthetic anomalies (0.1%, 1%, 5% and 10%) into naturally sampled light sensor data collected from a wireless sensor network. A five-fold cross validation method was implemented for training and testing with the results of each training set applied to all four test sets. The false positive and false negative rates, the traditional accuracy, and the geometric means were analyzed to determine the relationship of the number of anomalies assumed to occur in a test environment, the number of anomalies that actually occur in the environment, and the resulting performance of the anomaly detection algorithms.

Citation Download Citation

Leonard Lightfoot, Ellen Laubie, and Joseph Natarian "The effects of synthetically augmented training data on parameter tuning for anomaly detection algorithms", Proc. SPIE 8062, Defense Transformation and Net-Centric Systems 2011, 80620M (23 May 2011); https://doi.org/10.1117/12.883526

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available