Search

Software has become increasingly complex. It is very difficult to test software to the point where one can be confident that it is free of all faults. Therefore when used in applications, where high reliability is required, fault tolerance is necessary. We cover in this chapter the major approaches that have been developed to ensure fault tolerance in software. The topics covered include wrapping software to make it less failure-prone, the use of rejuvenation, N-version programming, and recovery blocks; software reliability models; exception handling, and fault-tolerant remote procedure calls.

Even a single data bit fault in an encryption or decryption circuit can spread quickly, and result in a totally scrambled output. Furthermore, deliberately injecting faults into a cryptographic device and observing the corresponding outputs can help an attacker gain access to the secret key. For these reasons, fault detection is needed in cryptography. We start this chapter with a brief overview of symmetric key and public key cryptography, and describe the fault injection attacks that can be mounted against them. We then present techniques that can be used to detect the injected faults in an attempt to foil such attacks.

This chapter starts with a discussion of how to express network resilience: both graph-theoretic and computer-network measures are considered. Common network topologies, such as multistage, crossbar, hypercube, mesh, and loop are described. Algorithms for fault-tolerant routing are presented. The chapter concludes with a discussion of on-chip, and of wireless sensor, networks.