Selection, Combination, and Evaluation of Effective Software Sensors for Detecting Abnormal Usage of Computers Running Windows NT/2000

Abstract

Intrusion-detection systems (IDS) can either: (a) look for known attack patterns, or (b) be adaptive software that is smart enough to monitor and learn how the system is supposed to work under normal operation versus how it works when misuse is occurring. They used approach: (b) in this project. Specifically, they empirically determined which sets of fine-grained system measurements are the most effective at distinguishing usage by the assigned user of a given computer from misusage by other insiders within an organization. In this project, they have made significant advances toward creating an IDS that requires few CPU cycles (less than 1 percent), produces few false alarms (less than one per day), and detects most intrusions quickly (about 95 percent within 5 minutes). The algorithm that was developed measures over 200 Windows 2000 properties every second, and creates about 1500 features out of them. During a machine-learning training phase, the algorithm learns how to weight these 1500 features in order to accurately characterize the particular behavior of each user-each user gets his or her own set of feature weights. Following training, every second all of the features vote as to whether or not it seems like an intrusion is occurring. The weighted votes for and against an intrusion are compared, and if there is enough evidence, an alarm is raised.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Apr 01, 2002
Accession Number
ADA406316

Entities

People

  • Jude Shavlik

Tags

Communities of Interest

  • Autonomy
  • Cyber
  • Electronic Warfare
  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Air Force Research Laboratories
  • Algorithms
  • Application Protocols
  • Basic Programming Language
  • Computer Science
  • Computers
  • Cybersecurity
  • Detection
  • Detectors
  • False Alarms
  • Intrusion
  • Intrusion Detection
  • Intrusion Detection Systems
  • Intrusion Detectors
  • Machine Learning
  • Operating Systems
  • Warning Systems

Fields of Study

  • Computer science

Readers

  • Cybersecurity.
  • Neural Network Machine Learning.
  • Sensor Fusion and Tracking Systems.

Technology Areas

  • AI & ML
  • AI & ML - Neural Networks