Algorithms for Efficient Intelligence Collection
Abstract
Modern intelligence techniques have drastically increased the rate at which communications data can be intercepted for analysis. This increased ability to collect data, coupled with the growing use of cell phones, SMS messaging, and email as methods of information sharing, means that intelligence agencies face a potentially overwhelming volume of intelligence data. The intelligence cycle describes the process by which intelligence data is collected, processed, and evaluated. In this thesis, we focus on the processing stage, where an intelligence processor screens the data, considering the information's reliability, validity, and relevance. This processing stage often requires human involvement to forward relevant intelligence data to analysts, and it is often time critical. The processor faces an information selection problem, and must decide which pieces of information to screen and in what order, to maximize the amount of useful data collected. We implement a mathematical model to handle the information selection problem, and develop a software library to allow for the testing of different heuristic screening algorithms on a variety of intercepted intelligence network structures. The software consists of the following main components: GraphBuilder, MapBuilder, and heuristic algorithms. Using our software, we analyze the performance of several screening strategies on a variety of representative intercepted intelligence networks, which we construct using real-world data sets. We show that the model consistently out-performs more naive approaches on networks with clusters of relevant sources. We also highlight the importance of exploration in robust screening strategies.
Document Details
- Document Type
- Technical Report
- Publication Date
- Sep 01, 2013
- Accession Number
- ADA590074
Entities
People
- Duncan R. Ellis
Organizations
- Naval Postgraduate School