Sometimes it’s up to the quality engineers, QA and testers to do all kinds of application security testing – especially in smaller firms, where the role boundaries aren’t that specific. In bigger companies it’s more often seen that it’s up to the security people to do security – and it is a whole world on it’s own. Yet being quality advocates we must also involve ourselves in security testing.
At work I’m currently anchored in the security team, as I lead testing of “all the things”. Including leading testing activities around security requirements and performing security tests. While my security colleagues set and interpret the security requirements, my role is to make sure it’s tested in some shape and form. And that point I’m sure many testing people can see themselves in. It might not be their specialized field – but they need to be versatile enough to have it on the list somehow.
October 2026 has indeed been a cyber awareness month as Marcus Hutchins recently pointed out. The PHP2shell exploit, the eMudra certificate pin incident, PWC data beach.. and the list goes on. While much have been talked about the technical details of what happened, how it could happen and how to fix it – let me address my testing and security audience and high-light how to move forward with four lessons learned.
1. The probability is both UP, but also less important
Classic risk- based testing (and risk management) will tell you create a matrix of business risk times probability. And as Katja Obring points out “Risk is likelihood times impact, and the age of the flaw changes neither….What changed this week is likelihood“. Even an old vulnerability can be exposed on short notice. If we have to address all vulnerabilities, perhaps probability is not the driver. anymore.
It reminds me of the challenges of the risk model it self. When you read the research from Daniel Kahneman & Amos Tversky, the classic model of “Degree of Risk = Probability x Impact” is flawed. Dr. Warren Black writes: “Human judgment is littered with natural imperfections such as subjectivity, variance, and bias, which they termed “Noise.” Too much Noise clutters and confuses the decision-making process, leading to major systemic vulnerabilities in all future actions.”
Over time perhaps we will se a better model for managing risks and balancing risk-based testing. I would prefer a way to factor first the customer importance (as before), but besides probability some representation of the feasibility (or maturity) of the risk. On one side open exposure on the global Internet (PHP2shell), then exploiting libraries which requires a little more access and context. And on the very far end niche issues in closed systems.
2. Insist that good (security) hygiene is important
It’s OK to police good quality practices and security hygiene – if no-one else is taking that role, pick it up. Insist that “Definition of Done” and “Definition of Ready” are there for a reason. Insist that passwords and pin codes are strong and stored safely. Insist that you have a “bill of material”, so that you know which versions your sites are running and that thy are up to date on alle the frameworks. Believe me when I say, it’s way cheaper and faster to update along the way, than to jump from PHP6 to PHP8 (or similar). It’s not about technical debt, it’s more a matter of being diligent on behalf of the company and stakeholders.
A key concept in the IT security space is the mnemonic “CIA” of the Confidentiality, Integrity, and Availability of a system [https://en.wikipedia.org/wiki/CIA_triad] – not the other CIA… The mnemonic helps identify risks to the system in each theme and helps organizations construct “controls” or requirements around mitigating those risks.
- Confidentiality: That things are not shared to unauthorized people or systems
- Integrity: That things are not modified, and that modifications are credible
- Availability: That the system is available when needed, and can recover from being unavailable.
All of these things can be tested – and all people looking into to quality (including security) should be involved in testing it.
A key trick is to communicate clearly to your customers and end-users, also those not affected. A local telephone company had a breach of private users, yet they also informed business users. Because the business users have leadership and oversight committees and auditors that will call the CEO and ask if they are in control. And that is your cue, help your manager to tell the CEO that they are in control and have a confident and trained plan.
3. Test and automate security requirements
All you company security policies and run-books are like backups. If you don’t use them you, they are worthless. And then the stakeholders cry. So what would they rather – A) have week of disaster recovery testing once a year or scrambling before the audits. Or B) build in quality and security requirement testing from the start? GRC engineering should be the next shift-left – the truth is out there.
We can already add engineering practices to the mechanics of confirming security requirements -and it’s not far from Quality Engineering.
4. Be part of incident response teams
The MotaVerse “This week in quality, Oct 23rd 2026” tells how quality professionals are involved in patching community sites (applying the actual fixes) and being part of incident response teams. One quality professional participated in the response team to monitor for the fix and push relevant PR’s. While it was a nice bonding exercise, it could surely help to have a run-book in place and a training in it. Recently at a conference I met a Quality Director who was also one of three on-call incident response and escalation manager.
Last year I was part of a “business continuity test” for a public sector economy solution, where we gathered all the key account people and simulated the site breaking. It was absolutely stil a formal test, with a test plan, test cases and bug tracker. Unfortunately the participants could not perform under even this simulated pressure, so the exercise eventually had to be rerun. With the directors and key account staff all in the same room. And they would rather have been doing something else, yet here again the bonding and training will pay-off for when the real crisis emerges.

2 responses to “Four things QE can learn from recent AI exploits”
Ministry of testing newsletter, October 30, 2026
https://www.linkedin.com/pulse/38-lessons-from-building-ai-products-myths-trouble-gug5e/
Issue #236 : Software Testing Notes
https://softwaretestingnotes.substack.com/p/issue-236-software-testing-notes