A machine learning approach to security improvement in mobile communication.
Item
-
Title
-
A machine learning approach to security improvement in mobile communication.
-
Identifier
-
AAI3187420
-
identifier
-
3187420
-
Creator
-
Sharif, Hooshang F.
-
Contributor
-
Adviser: Michael Conner
-
Date
-
2005
-
Language
-
English
-
Publisher
-
City University of New York.
-
Subject
-
Engineering, Electronics and Electrical
-
Abstract
-
One of the challenges for today's mobile communications engineers is designing networks that are secure and reliable. Over time, these networks have become increasingly complex. There are situations where a network's functionality and reliability must be maintained under particularly adverse circumstances. For example, in a modern combat environment, high jamming noise would necessitate signal transmission at high energies and under severe bandwidth constraints. Accordingly, there needs to be a significant reduction to the system's allocation of resources, such as bandwidth, that are dedicated to vulnerability and intrusion detection.;Conventional methods of vulnerability detection are often criticized for their predictability as well as excessive need for resource-allocation. Software agents, however, can be implemented to continuously monitor the network and apply the necessary resources in order to maintain the optimal security status.;In this dissertation, we have addressed the vital role of improving security and reducing vulnerability in a wireless network in the presence of the bandwidth limitation that may exist in a combat environment. We introduce methods that allow learning agents to anticipate vulnerabilities that are likely to occur in the network. Acquiring the knowledge for statistically correct anticipation will result in optimized use of bandwidth as well as other resources that are dedicated to security.;Although the main focus of this work is algorithms developed based on reinforcement learning, we have also developed alternative algorithms based on evolutionary computation methods for comparison.;It is the ability to learn the dynamics of a network environment in a continual and adaptive way that prompted us to utilize machine learning ideas as framework for developing our algorithms. We believe, however, that the methods introduced in this work can be applied to all mobile communication environments in general.
-
Type
-
dissertation
-
Source
-
PQT Legacy CUNY.xlsx
-
degree
-
Ph.D.
