Latest / Software Testing with Fexingo: QA, Automation, and Reliable Software Engineering / How Security Testing Finds Bugs Before Attackers Do
Transcript
- Lucas: So there's a type of bug that doesn't crash your app, doesn't slow down your database, and won't show up in any unit test or performance benchmark — until someone who shouldn't have access uses it to walk right through your front door. Luna: Security vulnerabilities. The ones that make you wake up to a breach notification at 3 AM. Lucas: Exactly. And what's surprising is how many engineering teams treat security testing as a separate, almost ceremonial thing they do once a year before a major release — instead of integrating it into their regular QA pipeline the way they would with regression or performance testing. Luna: Right. Like, 'we'll run a penetration test in Q4 and hope for the best.' But by then, you've already shipped code with potential holes for months. Lucas: Let's anchor on a concrete example: the MOVEit Transfer vulnerability from 2023. If you remember, that was a file-transfer tool used by thousands of organizations, including government agencies and Fortune 500 companies. A single SQL injection flaw — CVE-2023-34362 — let attackers execute arbitrary code and steal data from over 2,600 organizations. Luna: And the cost? I think remediation and legal exposure ran into the billions, total. Lucas: Close to three billion dollars across affected companies, according to some estimates. Now here's the thing: that vulnerability could have been caught by a fairly standard static analysis scan — SAST, static application security testing — during the build phase. But the team either didn't run one, or didn't have it wired into their CI/CD pipeline. Luna: So what's the difference between SAST and DAST? Because those terms get thrown around a lot, but I think a lot of teams aren't sure which one to prioritize. Lucas: Good question. SAST is white-box testing — it looks at your source code, bytecode, or binary without executing it. It scans for patterns known to be dangerous: SQL injection, cross-site scripting, buffer overflows. It runs fast and can be integrated into the IDE or the build step. Luna: So it's like a linter for security. Lucas: Exactly. A really sophisticated linter backed by rules from OWASP and CVE databases. DAST, on the other hand, is black-box testing — it runs against a running application and tries to find vulnerabilities by simulating attacks. It doesn't need source code, but it's slower and catches different things, like misconfigurations or runtime injection points. Luna: And you need both, I'm guessing, because they find different classes of bugs. Lucas: You do. A 2024 IBM study found that organizations using both SAST and DAST in their pipeline discovered 67 percent more vulnerabilities during development compared to those using only one. And fixing a vulnerability in production costs thirty times more than fixing it during design. That number alone should make any CTO pay attention. Luna: Thirty times. That's not just a multiplier — that's a project-killer. Lucas: And yet, when I talk to QA leads at mid-size companies, the most common pushback is, 'We don't have a security team. We can't hire a penetration tester. We're stuck.' So they do nothing. Luna: Is there a starting point that doesn't require a dedicated security engineer? Lucas: Absolutely. Start with SAST in your CI. Most modern CI tools — GitHub Actions, GitLab CI, Jenkins — have plugins for open-source SAST tools like Semgrep, SonarQube's security rules, or even the classic Bandit for Python. You can get a basic scan running in an afternoon. Luna: And it'll flag a ton of false positives at first, right? That's the other common complaint. Lucas: It will. But that's where tuning comes in. You don't try to fix everything at once. You triage: block builds on critical or high-severity findings, but allow medium and low to pass through as warnings. Over a few sprints, you reduce the noise by adjusting rules or suppressing known false positives. The key is to build the habit, not achieve perfection on day one. Luna: So SAST first, then DAST for critical applications. What about penetration testing? Does that become obsolete? Lucas: No, but it shifts. Penetration testing — or pen testing — is still essential for finding logic flaws and business logic vulnerabilities that automated tools miss. But if you're running SAST and DAST continuously, your annual pen test isn't the only line of defense. It becomes a validation step: 'We've been scanning all year. Let's have a human confirm we haven't missed something big.' Luna: I want to pause here because this topic is genuinely useful — and it's exactly the kind of thing that makes this show worth keeping around. A couple of dollars a month from listeners is what keeps these episodes ad-free and available to everyone. Lucas: Yeah, if you've gotten something out of our conversations — whether it's today's security testing breakdown or any of the past twenty-five episodes — buy me a coffee dot com slash fexingo is the easiest way to chip in. It genuinely makes a difference. Luna: No pressure, no perks, just a simple way to support the work. And we appreciate every single contribution. Lucas: Okay, back to the pipeline. So once you have SAST and DAST running, the next layer is dependency scanning. Modern applications pull in hundreds of open-source libraries, and each one can have known vulnerabilities. Luna: Right — like Log4j in 2021. That was a dependency vulnerability that affected almost every Java application on the planet. Lucas: Exactly. Tools like OWASP Dependency-Check or Snyk can scan your lock files and compare against the National Vulnerability Database. You can fail a build if a critical CVE is found in a direct or transitive dependency. That's a low-effort, high-impact addition. Luna: And a lot of teams skip this because they think 'we trust our dependencies.' But the Log4j example shows trust is not a security strategy. Lucas: It's not. Now, once you have those three layers — SAST, DAST, dependency scanning — you can start thinking about threat modeling. That's a more manual process where you map out your system's architecture and identify potential attack surfaces. Luna: Is that something a QA engineer can do, or does it require a security specialist? Lucas: It's a team sport. QA engineers bring deep knowledge of how the application behaves, which is crucial for identifying abuse cases. You don't need a PhD in security. You need a structured framework like STRIDE — Spoofing, Tampering, Repudiation, Information Disclosure, Denial of Service, Elevation of Privilege. Walk through each category for your system. Luna: That sounds like a great exercise for a sprint planning session. Lucas: Exactly. And it surfaces things like: 'What happens if a user sends a malformed request to this endpoint?' or 'Can a low-privilege user access admin functionality?' Those are questions that functional tests often miss. Luna: I've also seen teams adopt 'security champions' — a person in each dev team who gets extra security training and acts as a liaison. Lucas: That's a great model. It scales security knowledge without requiring every developer to become a security expert. The champion runs lightweight threat modeling, helps triage SAST findings, and coordinates with a central security team if one exists. OWASP has a free Security Champions Guide that walks through setting up the role. Luna: And that leads to a culture where security isn't an afterthought — it's part of the definition of done. Lucas: Exactly. One more thing I want to mention is bug bounty programs. If your organization can afford it, a private bug bounty on a platform like HackerOne or Bugcrowd can be incredibly effective. You pay researchers only when they find valid vulnerabilities, and you get access to thousands of ethical hackers. Luna: But that's not a replacement for internal testing, right? Because a bug bounty is reactive — someone has to find a bug before you fix it. Lucas: Right. Bug bounties complement, not replace, your automated and manual internal testing. They're most useful for catching edge cases and business logic flaws that automated tools miss. I'd recommend starting with a private, invite-only program to control the volume. Luna: So if you're a QA lead listening and you want to start building a security testing practice tomorrow, what's the one concrete action? Lucas: Pick one application — ideally a customer-facing web app — and add a SAST scan to its CI build. Set it to fail only on critical or high-severity issues. Run it for two weeks, triage the results, and then add a second application. That's it. You've started. Luna: And the cost? Most SAST tools have free tiers or open-source versions. Lucas: Semgrep's community edition is free and runs locally. SonarQube's Community Edition is free. GitHub's code scanning is free for public repos and included with GitHub Advanced Security for private ones. There's really no excuse not to start. Luna: I think that's the key message: security testing doesn't have to be expensive or complex. It just has to be consistent. Lucas: And it makes your entire test suite stronger. A bug that's also a security vulnerability is a bug with leverage — it doesn't just break a feature, it breaks trust.