FPGA Viruses

Abstract

Programmable logic is widely used, for applications ranging from field-upgradable subsystems to advanced uses such as reconfigurable computing platforms. Users can thus implement algorithms which are largely executed by a general-purpose CPU, but may be selectively accelerated with special purpose hardware. In this paper, we show that programmable logic devices unfortunately open another avenue for malicious users to implement the hardware analogue of a computer virus. We begin with an outline of the general properties of FPGAs that create risks. We then explain how to exploit these risks, and demonstrate through experiments that they are exploitable even in the absence of detailed layout information. We prove our point by demonstrating the first known FPGA virus and its effect on the current absorbed by the device, namely that the device is destroyed. We close by outlining possible methods of defense and point out the similarities and differences between FPGA and software viruses.

This work was supported by DARPA under Contracts #DABT63-95-C-0073, #N66001-96-C-852 and #MDA972-95-1-0013, with additional support from the Hewlett-Packard and Intel Corporations.