Publications

This reports describes the successful extension of artificial immune systems from the domain of computer security to the domain of real time control systems for robotic vehicles. A biologically-inspired computer immune system was added to the control system of two different mobile robots. As an additional layer in a multi-layered approach, the immune system is complementary to traditional error detection and error handling techniques. This can be thought of as biologically-inspired defense in depth. We demonstrated an immune system can be added with very little application developer effort, resulting in little to no performance impact. The methods described here are extensible to any system that processes a sequence of data through a software interface.

Many governmental and corporate organizations are interested in tracking materials and/or information through a network. Often, as in the case of the U.S. Customs Service, the traffic is recorded as transactions through a large number of checkpoints with a correspondingly complex network. These networks will contain large numbers of uninteresting transactions that act as noise to conceal the chains of transactions of interest, such as drug trafficking. We are interested in finding significant paths in transaction data containing high noise levels, which tend to make traditional graph visualization methods complex and hard to understand. This paper covers the evolution of a series of graphing methods designed to assist in this search for paths-from 1-D to 2-D to 3-D and beyond.