Dev.to Security 🔐 Cybersecurity 👁 0 📖 5 min read

The IT Side of a Pen Test Week, a Client-Side Runbook

A pen tester at a company of 30 to 80 people usually works for 3 to 5 days, and the prep around those days takes closer to 4 weeks. This is the runbook I use from the client side for the IT part of that month, written fo

A pen tester at a company of 30 to 80 people usually works for 3 to 5 days, and the prep around those days takes closer to 4 weeks. This is the runbook I use from the client side for the IT part of that month, written for whoever on staff ends up owning it.
Defending against the tester isn't on the list. What is on it is knowing which traffic is theirs, not changing the network mid-test, having someone who can stop things, and protecting the old servers that can't take much load.
The problem
The report you pay for describes one week of your network. Push a firewall change on day 2, or let the tester get blocked on day 1, and it ends up describing a week that didn't really happen, which nobody on your side can do much with.
NIST SP 800-115 calls planning the first phase of a security test (it comes before any testing traffic is sent), and in my experience nearly all of the planning below lands on IT.

  1. Get the tester's source addresses into the scope letter Along with the systems in scope by IP address, domain and application, the dates and hours, and banned actions such as denial of service or copying real customer records, the scope letter we use lists the addresses the tester will work from. That last item is for the logs. After the test I go through the client's firewall logs with that address list next to them, and anything from those addresses is the tester's, while anything else that hit the edge that week gets looked at on its own. Without the list we'd be going back and forth on every odd connection.
  2. Decide what your monitoring provider is told I don't think there's a clean answer here, and the options I weigh with clients look like this. Tell nobody at the provider, and you find out whether they'd have caught a real intruder. Tell their whole analyst team, and the test stops telling you anything about detection. Tell nobody and get blocked the morning of day 1, and on a 5-day test that's 20% of what you paid for gone. My compromise is one senior person at the provider who knows, and nobody on their analyst floor. If you use (or are about to sign with) a managed detection and response provider, ask before signing how they'd treat alerts from the tester's addresses. As for your own staff, the CIS Controls v8.1 guidance on penetration testing says only a few people in an organization should know, with one person named as the main contact if something goes wrong. On my projects that's the owner, the IT lead, and whichever manager has a team that might notice odd logins (plus HR, in writing, before the first phishing email is sent if phishing is in scope).
  3. Patch the edge before the tester arrives CISA added 321 entries to its Known Exploited Vulnerabilities catalog between November 17, 2025 and November 13, 2026. Of those, 64 were in firewall and remote access products from makers like Cisco, Fortinet, Ivanti, SonicWall and Citrix, and Microsoft products accounted for another 42. Since that's the gear attackers go after first, it's also the gear my clients patch first. On our countdown that happens 10 days before kickoff. Check those entries against your own firewall, VPN and Microsoft products, update firmware, turn on MFA for every admin and remote access account, change default passwords, and leave the rest alone. I know some prep guides say to fix everything first. I don't agree, since then you're paying a tester to confirm your scan results. The snag I see on this step is a firmware update needing a maintenance window nobody booked, so I ask for that window to be booked when the letter is signed.
  4. Protect the fragile machines Windows Server 2012 R2 is the usual one. Extended support ended in October 2023 and the last paid security update year ended October 14, 2026, and I still find those machines in Dallas server closets running payroll or a line-of-business database. They go into the scope letter with a note to test gently or skip them, along with anything else that can't take much load. The day before testing, take a fresh backup of them. A week out (day 7 on our countdown), do one real restore of a file share or virtual machine and write down how long it took. CISA's #StopRansomware Guide says to test backup procedures regularly anyway, and a console that says "success" hasn't shown you the data comes back.
  5. Staff the window I want 2 names and 2 cell numbers on the client side for the test window. One of them can call a pause, and the other can restart a server in the evening if one goes down. You don't need a security analyst for this. BLS counts 7,080 information security analysts in the Dallas-Fort Worth metro in May 2025 at a mean of $133,790 a year, and a 60-person company can't usually carry that salary, so the 2 names on my contact sheet tend to be someone from the IT provider and the office manager or controller. There's a stop word as well, and testing stops when anyone says it. The letter also says what happens if the tester sees signs of a real attacker. In ours, testing pauses and the tester calls the named contact before doing anything else.
  6. Freeze changes Hold off on nonurgent changes for the week. A firewall rule somebody pushes on Wednesday can leave Thursday's results unreadable, since then you can't say which configuration the tester saw.
  7. Handle the report like a credential CIS warns the report can work as step-by-step directions for breaking in. I keep copies to the 2 or 3 people doing the fixes, and one goes in the client's security records. When testing ends on a Friday, the readout call goes on the calendar for the next week, 2 or 3 days of IT time get blocked right after it for the critical findings, and a retest is set about 30 days out. Gotchas On the countdown I send Dallas clients about a month out, these are the snags that come up at the IT steps. Staffing (step 9 on the full countdown) goes wrong when the only admin has vacation booked that week. On step 6 it's a firmware update that needs a maintenance window nobody booked. At step 8 the SOC blocks the tester on day 1, and the test stalls. For step 7, a backup job that reports success turns out never to have been restored. And at step 10 the report lands, then sits in an inbox for a quarter. Takeaway Looking back over that list, very little of it is security work in the usual sense. It's logistics, and when it's done the report you get back should describe the network you actually run rather than a week of blocked traffic and half-made changes. The full 10-step version, including the Texas law behind the scope letter and vendor approvals, is in our guide on how to prepare for a penetration test in Dallas.
📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.