Software Streams: Big Data Challenges in Dynamic Program Analysis

Abstract

Dynamic program analysis encompasses the development of techniques and tools for analyzing computer software by exploiting information gathered from a program at runtime. The impressive amounts of data collected by dynamic analysis tools require efficient indexing and compression schemes, as well as on-line algorithmic techniques for mining relevant information on-the-fly in order to identify frequent events, hidden software patterns, or undesirable behaviors corresponding to bugs, malware, or intrusions. The paper explores how recent results in algorithmic theory for data-intensive scenarios can be applied to the design and implementation of dynamic program analysis tools, focusing on two important techniques: sampling and streaming.