The SCIF Drone Threat: Rethinking RF Risk on Finished SCIFs and SAPFs
For years I've made the same half-joke in classes and design reviews: man bars won't stop a drone. Man bars do their job. They keep a person from crawling through a duct into the secure space. They do nothing for RF energy. That takes a waveguide.
Here's why the joke stopped being funny. A duct with man bars and no waveguide has to end somewhere, and usually it ends at a rooftop louver. That louver is exactly where a drone would land. I'm not talking about a drone watching a SCIF get built. I'm talking about the finished, accredited space, and what a cheap drone can do to it once it's operating.
The assumption underneath RF risk decisions
When a CTTA decides how much TEMPEST and RF protection a facility needs, one of the biggest factors is inspectable space. The national definition describes it as the three-dimensional space around equipment processing classified information within which TEMPEST exploitation is not considered practical, or where there is legal authority to identify and remove a potential TEMPEST exploitation.
That definition has two halves, and drones break both of them.
The first half is proximity. Could someone get close enough to collect? For a facility inside a fenced installation, with guards, patrols, and open ground around it, the answer has long been "not practically," because getting close meant walking, driving, or renting the office next door.
The second half is legal authority. If someone did get close, could we do something about it? Against a person, usually yes. Against a drone, usually no. Federal counter-drone mitigation authority is narrow by statute. 10 U.S.C. 130i covers specific DoD facilities, 6 U.S.C. 124n covers specific DHS and DOJ missions, and DOE has its own separate authority. Recent legislation has started to widen those authorities, but a contractor SCIF, and most agency space outside the covered facilities, still has no legal authority to jam, spoof, or take down a drone. FCC rules prohibit jamming outright.
So for most facilities, both conditions behind the inspectable space assumption fail at the same time.
Why the SCIF drone threat is different
A drone is a delivery system. The payload is whatever someone wants it to be: a camera, a radio receiver, a recorder, an explosive, or something designed to disrupt electronics. In Ukraine we've watched the same basic airframes used for reconnaissance, as loitering munitions, and as one-way attack vehicles, built quickly and cheaply from commercial parts.
Two developments from that war matter here. Both sides now reportedly field drones guided by a thin fiber-optic cable that unspools in flight, with no radio link to jam or detect, and one Ukrainian strike drone is reported to wait in ambush for up to 48 hours. Ukrainian forces have also reportedly used AI software that compares live camera video against an onboard map, so a drone can navigate by sight with no GPS at all.
Put those together and you get something that can fly to a building without broadcasting anything, find its spot visually, and sit there. Most counter-drone detection leans on the radio link or GPS behavior. A drone that uses neither is hardest to find on the way in and on the way out. The fiber matters for a second reason, too: it's a high-bandwidth path back to the operator with no RF signature at all.
Three things a drone can do to a finished facility
What follows is my own assessment as an ICD 705 practitioner and security professional, not a list of reported incidents. Each of these is within reach of a cheap drone today.
Sit at the weak spot and collect. Every shielded space has places where performance is lowest, and windows are usually the first. Agencies like to keep their windows, and shielded glazing has come a long way, but it often underperforms the surrounding wall assembly, especially at higher frequencies and especially at the frame, where the bond and gasket matter more than the glass. Skylights, rooftop penetrations, and any spot where the shielding was never quite right fall in the same category. And a window isn't only an RF problem. A drone on the ledge can carry a camera looking in, or a laser or vibration sensor reading speech off the glass. ICD 705 already requires visual and acoustic treatment of windows, but those treatments assume the observer is at a distance. A drone on the ledge collapses that distance for RF, acoustic, and visual all at once.
To be fair, close-in collection isn't trivial. Modern low-voltage electronics radiate less than older gear, and turning a weak signal into something useful takes real skill. But distance was the main thing keeping it impractical, and a drone removes the distance.
Leave something behind. A drone doesn't have to stay. It can set a small collection device on a roof, a parapet, or a window ledge and leave. That device has two choices: store what it collects and wait to be picked up, or transmit, which creates an RF signature that monitoring could catch. Store-and-retrieve leaves very little to find. Rooftops and ledges are rarely part of anyone's routine security checks, so it's entirely possible no one would know.
Take the mission down instead. This is the one I don't hear anyone talking about. A drone can carry a high-power microwave or intentional electromagnetic interference device (HPM/IEMI, often loosely called EMP) and deliver it right to the building. A drone-portable source is far more likely to disrupt than to cause permanent damage, but disruption is enough. The attacker may never get a single classified signal out. They don't need to. If a mission-critical lab goes dark, the mission is down, and the cost to the attacker could be a drone that sells for a few hundred dollars plus whatever it's carrying.
Architectural shielding attenuates in both directions, so the walls do help. But that shielding was specified and tested to keep low-level emanations in, not to survive a high-energy pulse. The real exposure is conducted: power and signal lines that carry the energy straight past the shield. TEMPEST power filters aren't surge protection, and a high-energy transient can damage the filter itself. HPM sources also tend to operate in the low gigahertz range, which is where seams, gaskets, windows, and imperfect waveguides perform worst.
Facilities hardened against HEMP are built to a different standard altogether, MIL-STD-188-125-1, which calls for roughly 80 dB of shielding, surge protection at every point of entry, and pulse-injection verification. The Army Corps of Engineers just solicited a $150 million design contract (SAM.gov notice W9128F27RA016) covering both SCIFs and HEMP shielding, so some facilities are being built that way. Most secure spaces are not.
And an attacker doesn't have to touch the SCIF at all. Rooftop condensers, generators, transformers, and communications antennas are far easier targets than a shielded room. Lose cooling, and a server-heavy SAPF goes down quickly.
The balloon was the long-range version
In 2023 a Chinese surveillance balloon crossed the continental United States before it was shot down off the East Coast. NBC News reported that it collected mostly electronic signals and could send them back to Beijing in real time. In June 2023 the Pentagon said its assessment was that the balloon did not collect while over the United States.
Reporting conflicts on what it actually collected. But the intent was clear: signals collection over our facilities is a real objective. From that altitude, compromising emanations were never the realistic target. A drone sitting on your window ledge is a very different distance.
Where I've landed
The insider remains the biggest risk to any SCIF or SAPF. Nothing here changes that.
When I look at the protections we build, we do a good job on the physical side, and a good job with alarms, access control, and electronic security systems. RF has always been the trickier layer, and it's the one where drones change the math the most.
My view is this: outside of a space nested inside another building, ideally one with some shielding of its own, I don't think any above-ground facility can assume its inspectable space will hold against a small drone. That's true CONUS and OCONUS. A fence line and a guard force control who walks up to the building. They don't control what lands on the roof for twenty minutes and leaves, and in most cases they have no authority to stop it.
And given how fast drones have moved into warfare, I think it's more likely, not less, that they'll be pointed at SCIFs and SAPFs, whether to collect or to disable.
What I'd like teams to start asking
None of this means every facility needs to be fully hardened. That would be expensive and often unnecessary. It means the RF and TEMPEST conversation should account for the SCIF drone threat: something that can reach the outside of the building quietly, and for an attacker whose goal is to stop the mission rather than steal from it. Questions I think belong on the table:
Do the windows and skylights perform like the walls? For RF, acoustic, and visual protection, with an observer on the ledge rather than across the street. If not, can they be removed, upgraded, or kept away from where processing happens?
Where is processing relative to the exterior? Equipment against an exterior wall or under a roof deck is easiest to reach. Interior rooms buy distance for very little cost.
Are penetrations treated for RF, not just for people? Man bars and waveguides solve different problems. Both questions deserve an answer, especially where a duct ends at a rooftop louver.
Is the roof inside camera and IDS coverage? That's cheaper than adding the roof to patrol rounds, and it's how you'd find something left behind.
Are power and signal entry points protected against surges, not just filtered? That's where a pulse actually gets in.
If the lab went dark tomorrow, how long would it stay dark? Include the rooftop condensers, generators, and transformers. Protected spares and off-site backups are mission-continuity questions, not just accreditation questions.
Who has authority to act on a drone over this facility? If the answer is "no one on site," the plan should say who gets called and how fast they can respond.
Are drones and HPM in the facility's threat picture at all? If not, that's worth raising with the CTTA and the AO now, while the design can still change.
These decisions belong to the CTTA and the AO, and they should. My hope is that design and construction teams, and the security managers who run these spaces, bring the question to them early rather than finding out later that the threat picture has moved.
Let's talk about it
I don't have every answer here, and I'd rather this be a conversation than a lecture. If you're a CTTA, an AO, a security manager, or a designer who has started working through drone or HPM risk on a secure facility, I'd like to hear how you're thinking about it. And if your team wants a second set of eyes on the RF and TEMPEST approach in light of this, we're glad to help.
Comments