The Wall Was Twelve Inches Thick Until It Reached the Ceiling

We were on site at a data center in a remote part of South America that had been running for years, and it was as hard a target as we have assessed. Armed guards worked the site. The outer fence was ten feet of welded wire topped with concertina, with a hundred feet of open ground between it and the next fence inside. The building shell was twelve-inch fiber-reinforced concrete block. The doors were twelve-gauge steel with three-point locks. The windows were glass-clad polycarbonate thirteen-sixteenths of an inch thick. Every surface of that building had been built to keep a determined person out, and it did.
Just inside the front door was an outer alcove that housed the storage tanks for the fire suppression system. The ceiling in that alcove had never been finished. The twelve-inch wall between the alcove and the rest of the building stopped short of the roof deck instead of running all the way up to it. Our tester climbed onto one of the tanks, went over the top of the wall through the open ceiling and dropped into a bathroom on the other side. The wall that would have taken hours to breach was not there above a certain height. Nobody who worked in the building knew it, because nobody had ever checked that the building was finished the way the drawings said.
Buildings are secured to the drawing
Physical security is designed in layers. There is a perimeter, then the building shell, then the interior and then the room that matters. Each layer is meant to do two things. It slows an intruder down, and it gives the people protecting the site a chance to notice. A security designer specifies each layer on paper. The fence is this tall, the wall is this thick and the door is this gauge. A contractor builds it, an auditor confirms the controls exist and the site goes into service.
The problem is that the layers only work as a sequence. An intruder does not have to beat the strongest part of the building. They only have to find one path where every step is weak, and they can take as long as they want to find it. One unverified gap removes the delay of everything around it. The data center's concrete, steel and polycarbonate added nothing once there was a way over the wall that took thirty seconds.
That gap was not a design failure. The drawing showed a wall that ran from the floor slab to the roof deck. The failure was that nobody compared the finished building to the drawing. Audits check that controls exist. They rarely check that a control is continuous, meaning the wall actually reaches the deck, the door actually closes against its frame and the badge reader actually rejects the old cards.
Why the strong parts get the attention
Budgets go to the parts of security people can see. A heavy door, a new camera and a guard at the desk are easy to buy, easy to show a board and easy to audit. The things that defeat them are not on anyone's list. The space above a dropped ceiling belongs to facilities. The request-to-exit sensor above a door was installed so people can leave in a fire. The old proximity card support on a new badge reader was left on during a changeover and never turned off. None of those has an owner whose job is to ask whether it undoes the controls around it.
We see the same pattern on almost every physical test we run. The client's most expensive controls usually work. The way in runs through something that was installed for convenience, for safety or for a project that finished years ago.
What you can check yourself
Start above the ceiling. In any room that holds something you care about, lift a ceiling tile next to the wall of the secured room and look up. If the wall stops at the tile grid and the space above runs over the top, anyone in the next room can climb over it. Walls around server rooms, records rooms and executive suites should run from the floor slab to the deck above. In most buildings at least one of them does not.
Walk every secured door from the outside. Look for the gap at the top of the door and between a pair of doors. Then look on the inside for a request-to-exit sensor, abbreviated REX. That is the small motion sensor mounted above many secured doors that unlocks the door when someone walks up to leave. It has to be easy to trigger so nobody is trapped in a fire. If something pushed through the gap from outside can reach the space the sensor watches, the door opens itself for anyone who knows that. A latch you can see through the gap, a thumb turn close to the edge of a door and a lever handle above a gap under the door are all the same problem.
Read the model on your badge readers. Many sites moved to encrypted cards years ago and left the readers set to accept the old proximity cards so nobody was locked out during the switch. A proximity card can be copied from a few inches away, and a reader that still accepts one opens for a copied card no matter what employees carry. If your reader's model name says it supports proximity, find out whether that support is still turned on.
Watch your own building at 1000 on a weekday. Find the door the smokers use and see whether it gets propped. Then pick the alarm on your most important rear door and ask who answers it at 0200 and how long that takes. Ask for the measured time, not the contract. The first answer is usually a guess, and the measured one is usually longer.
Each of these takes an afternoon. Most organizations that try them find more than one, and the ones they find tend to lead to others nobody has looked at.
The book
Everything above comes out of a new handbook, Physical Penetration Testing. It is written for people coming into the field, and it walks through a full physical engagement from the first scoping meeting to the final report. It covers authorization and the rules of engagement, reconnaissance, perimeters, locks and the door tools that get past them, and badge cloning with the Proxmark3. It also covers alarms and cameras, social entry, covert entry and writing findings a client will act on. It includes the templates we use for the authorization letter, the target package, the operation plan and the closeout. The data center above is one of its chapter openings, and every chapter opens with a real engagement.
For a limited time the book is free to anyone who wants it. Email keith@redcellsecurity.org or send a direct message on LinkedIn or Facebook and we will send you a copy. It goes on sale on Amazon the first week of December.
It is Book 12 of The Red Cell Penetration Testing Handbooks, a series of fifteen short handbooks for people getting into penetration testing. The others cover foundations, reporting, project management, reconnaissance, network discovery, vulnerability validation, exploitation, web applications, passwords, Active Directory, wireless, AI, social engineering and cloud. They will come out on Amazon over the next few months.
What we do with this
We run physical penetration tests the way the book describes. We start from your real buildings, find the path an intruder would take and walk it up to the objective. We prove we got there and leave everything as we found it. The report shows every step of the path and the one or two changes that would have stopped it. It also names the controls that did their job, so you know what to keep.
For sites still being planned or renovated, we do target assessments that compare the building to its drawings before an intruder does. That is how a wall that stops at the ceiling gets found by a tester with a flashlight and not by someone standing in your server room.
We're based in Dallas, Texas, and we go wherever the problem is. Red Cell Security provides cybersecurity and physical security consulting to clients across the US and around the world.




Comments