Building Secure and Reliable Systems

Chapter 13. Testing Code

By Phil Ames and Franjo Ivančić

with Vera Haas and Jen Barnason

A reliable system is resilient to failures and meets its documented service level objectives, which may also include security guarantees. Robust software testing and analysis are useful aids in mitigating failure risks, and should be a particular focus during the project implementation phase.

In this chapter, we discuss several approaches to testing, including unit and integration testing. We also cover additional security deep-dive topics like static and dynamic program analysis and fuzz testing, which can help strengthen your software’s resilience against the inputs it encounters.

No matter how careful the engineers developing your software are, some mistakes and overlooked edge cases are inevitable. Unexpected input combinations may trigger data corruption or result in availability issues like the “Query of Death” example in Chapter 22 of the SRE book. Coding errors can cause security problems like buffer overflows and cross-site scripting vulnerabilities. Put simply, there are many ways software is prone to failure in the real world.

The techniques discussed in this chapter, used in different stages and contexts of software development, have a variety of cost–benefit profiles.¹ For example, fuzzing—sending random requests to a system—can help you harden that system in terms of both security and reliability. This technique can potentially help you catch information leaks and reduce serving ...