Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Security chaos engineering deliberately breaks a control in a measured way to find out whether detection, response, and the control itself behave the way the design documents claim.

What Is Chaos Engineering in Security?
Security chaos engineering borrows a practice from site reliability work and points it at defences. Rather than assuming a control functions, you break something deliberately and in a measured way, then watch whether detection, response, and the control itself behave as the design documents claim. The output is not a pass mark. It is a list of assumptions that turned out to be wrong while nobody was attacking you.
Netflix released Chaos Monkey in 2011 to terminate production instances at random, forcing engineers to build services that survived the loss of a node. The security variant came later: Aaron Rinehart built ChaoSlingr at UnitedHealth Group in 2017 to inject security faults into an AWS estate, and the approach was documented properly in Security Chaos Engineering by Kelly Shortridge and Rinehart in 2023. The underlying claim is the same in both cases. How a system behaves under failure is an empirical question, and the architecture diagram is not evidence.
An experiment states a hypothesis, changes one variable, and measures a defined steady state. A handful reliably produce surprises:
Each has a documented expected result written before the experiment runs. The value sits entirely in the gap between what you expected and what you observed.
Blast radius comes first. Define the smallest change that can still falsify the hypothesis, agree an abort condition, and name the person who can call a halt. Staging is the sensible place to rehearse, but staging rarely has the same identity model, traffic, or third-party integrations, so the interesting findings tend to appear only in production. Decide in advance whether to tell the SOC. Telling them tests the control; not telling them tests the people, and conflating the two makes the result unreadable.
Tooling matters less than discipline. AWS Fault Injection Service, Chaos Mesh, Chaos Toolkit, and Gremlin all automate the mechanics, but a scheduled experiment described in a one-page document with a hypothesis, a metric, and a result is worth more than an unmaintained automation suite.
A penetration test asks whether an attacker can reach something valuable. A chaos experiment asks whether a specific control keeps working when conditions shift, and it answers that question repeatedly and cheaply. A tabletop exercise tests human decisions against a hypothetical; a chaos experiment tests the real system and its real alert routing. The three complement each other, and findings from each should feed security regression testing so that a control you fixed once stays fixed.
For financial firms the regulatory framing already exists. The FCA and Bank of England operational resilience rules took effect on 31 March 2022, with firms expected to be operating within their impact tolerances by 31 March 2025. They require organisations to identify important business services, set tolerances for disruption, and test their ability to remain within them under severe but plausible scenarios. Evidence that a security control degrades predictably, and that someone notices, is exactly the kind of testing those rules expect. EU financial entities face comparable duties under DORA, and organisations in scope of the NIS Regulations face similar questions about whether their controls have ever been verified in anger.
At Privia Security, we design and run this kind of controlled failure testing as part of our cyber exercise services.
It can be, provided the blast radius is bounded and the abort condition is agreed before you start. Begin with read-only observations and reversible changes on a single host or namespace, run during working hours when the people who own the system are available, and never chain two experiments together. If an experiment cannot be undone within minutes, it is not an experiment yet.
A red team operation simulates an adversary pursuing an objective, and success means reaching that objective undetected. A chaos experiment tests one control against one hypothesis, with the expected outcome written down in advance. Red teaming tells you whether your defences hold against a determined opponent; chaos engineering tells you which individual assumptions are already false, which is cheaper to learn and easier to fix.
Pick a steady-state metric that reflects the control working, not the system being up. Useful ones include mean time to detect, whether an alert reached a human, whether the admission or blocking decision was enforced rather than logged, and how long remediation took. Record the result even when nothing happened, because a silent failure is the most valuable finding in this kind of work.
No. Most organisations get their first dozen findings from manual experiments documented in a shared page: change one setting, observe, revert, write it down. Platforms such as AWS Fault Injection Service or Chaos Mesh become worthwhile once experiments need to run on a schedule, across many accounts, or with automatic rollback. Buying the tooling first usually produces an unused platform rather than findings.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.