Malware resistant networking using system diversity

Abstract

In the biological world, species are known to ensure survival by introducing variations among individuals of the species, some of which may have superior survival characteristics in the event of certain environmental changes. While not precisely equivalent, networked computer systems also must 'survive' in a rapidly changing environment. In particular, as with its biological analog there are predators in the network world that must be thwarted.This paper shows that in the macro-, as well as in the microsenses, system diversity is an important tool for network survivability. A malware attack on the Internet can have serious implications for many nodes but is unlikely to cross operating system or server application boundaries and thus to entirely disrupt the heterogeneous Internet.Individual networks that attach to the Internet can be developed using layers of servers, each with a different hardware and/or software basis, to isolate core servers from networked malefactors. Such a topology requires an attacker to employ a wide range of exploits, different ones tailored to each system at each protective layer of systems. This paper demonstrates how the layering technique protects networked systems.

Improving sensor network immunity under worm attacks: A software diversity approach

“…In parallel to our work, [20] illustrates that a mal-packet with only specially crafted data can exploit memory-related vulnerabilities and utilize existing application codes in a sensor to propagate itself without disrupting sensors' functionality. Software Diversity Inspired by diversity, an important source of robustness in biological systems, software diversity in computer systems [17] as well as computer networking [27,9] bears a lot of attention recently. A variety of randomization techniques [19] have been proposed to enhance the intrusion resistance of computer systems by increasing software complexity without degrading functionalities and performance.…”

“…While the general idea of software diversity is not new and it has been applied to wired networks [17,27,9], its application to sensor networks faces unique challenges due to high node density and sensor deployment error. High node density implies that it is unrealistic to assign a different version to each node; deployment errors could lead to the potential danger that sensor nodes with the same version of code become neighbors (connected) after they are deployed into the field.…”