Forty minutes
At 7:02 on the morning of December 7, 1941, two privates running a mobile radar at Opana Point on the northern tip of Oahu saw the largest return either of them had ever seen. A formation of aircraft was coming in from the north, about 130 miles out. They called it in to the information center at Fort Shafter.
The officer on duty was a lieutenant who had been to the center once before. He had little training, no supervision, and no staff. He knew a flight of B-17s was due from California that morning, and he told the privates not to worry about it. The attack began about forty minutes later.
The radar worked. The sensor did its job at a range that should have bought the fleet the better part of an hour. What failed was everything after the sensor: the picture, the person interpreting it, the authority to act, and the path from a warning to a response. Eighty-five years later, American installations are buying sensors again, and the same question applies.
The threat came home
In the week of March 9, security forces at Barksdale Air Force Base reported multiple waves of 12 to 15 drones over sensitive areas, including the flight line of a bomber base that stores nuclear weapons. The drones had non-commercial signal characteristics, long-range control links, and resistance to jamming. A briefing noted that their entries and exits seemed designed to keep the operators from being located, and that the operators may have been testing how the base would respond.
Northern Command had one counter-drone fly-away kit for the homeland at the time. Its commander said about a quarter of detected drones were being defeated, which was an improvement on almost none the year before. We wrote about the pattern of incursions over US bases in July. The pattern is persistent, deliberate, and cheap for the attacker.
The buying has started
Money is now flowing to the problem. On July 1, Joint Interagency Task Force 401 established a three-year, $500 million vehicle for its Domestic Shield program, and the first task order, worth $80.5 million, buys multi-sensor detection systems and counter-UAS radars for Air Force Global Strike Command sites. In May the task force named five installations for a directed-energy pilot: Fort Huachuca, Fort Bliss, Naval Base Kitsap, Grand Forks, and Whiteman. The FY27 request asks for $54.6 billion for the Defense Autonomous Warfare Group, up from $225.9 million this year, and counter-drone work sits inside that portfolio.
The task force's published guide emphasizes passive detection: radar, electro-optical and infrared cameras, and RF sensing that identifies drones by signal characteristics without reading the content of communications. That emphasis is sound. Detection comes first, and most installations have had too little of it.
Point defense is necessary and insufficient
In July the Mitchell Institute argued that point defense is necessary but insufficient against small drones. A single base can defend its own perimeter. It cannot see an operator probing five installations across a region in one night, and it cannot hand a track to its neighbor when the drone crosses the fence line.
The economics favor the attacker at every seam. A drone that costs a few thousand dollars forces the defender to spend minutes and sometimes a very expensive interceptor. We laid out the interceptor cost math in June. The defender can only close the gap by deciding faster and engaging with the cheapest effector that will work, and both depend on a shared picture.
Four seams
Installation defense has four seams where Opana-style failures happen. The first is between sensors. A base that buys radar from one vendor, RF detection from a second, and cameras from a third has three pictures, and a watch officer correlating them by eye. The second is between the picture and the authority to act. If the person who can authorize an engagement is at a regional operations center, the decision waits on a phone call and a working link.
The third seam runs between installations, and between the military and civil authorities. A drone crossing from one base's airspace into another's starts over as a new detection. Barksdale's response involved federal and local law enforcement, who were looking at the same sky with different tools. The fourth is between the event and the review. After a probing night, someone has to reconstruct what was seen, by which sensor, what was decided, and why, usually from screenshots and memory.
What the picture layer must do
Closing those seams takes a layer between the sensors and the effectors, and it has six jobs. It has to accept observations from any vendor's sensor in a common form, so adding a sensor adds detections, not another screen. It has to fuse them into tracks and fail closed, refusing reports it cannot verify, because an adversary who can spoof a track can send the response team to the wrong gate. It has to run at the installation and keep its delegated authority when the link to higher is cut.
It also has to share tracks across installations and with civil partners at the level each is allowed to see. It has to task effectors and response teams as assignments with an owner and a deadline, not as radio calls. And it has to record every observation, decision, and engagement, while retaining no more than the mission requires, which matters on a domestic installation where privacy rules apply.
How EdgeLance closes the seams
EdgeLance treats every sensor as a source of typed observations. Radar returns, RF detections, camera detections, and ADS-B reports all arrive in the same vocabulary and feed one multi-domain tracker. The tracker validates each report against the mission's schema, refuses backdated and downgrading inputs, keeps contested identities visible as separate tracks, and cannot be exhausted by a flood of conflicts. A new sensor joins the picture by producing observations, which is how the sensor fabric was designed from the start.
The picture runs on a forward node at the installation, with authority delegated by role. If the link to the regional center drops, the base keeps tracking and keeps its authority, and the record syncs back when the link returns. Tracks are shared across installations over the EdgeLance mesh, filtered by what each partner is cleared to receive.
Engagements run through the same tasking layer that EdgeLance uses for drones and ground teams. A confirmed threat becomes a task offered to the effectors and teams that can handle it. The best-placed one takes it on a lease, and the result comes back as a receipt. Every step lands in the mission record, and retention follows the mission's policy, so data that is no longer needed is destroyed on schedule while the record of decisions survives.
EdgeLance installs on base networks that never touch the internet, from one signed package, and works with whatever sensors and effectors the installation already owns. The base defense mission page shows the workflow end to end.
What installation commanders should ask for
Commanders and program offices can push every vendor toward a shared picture through requirements. Ask that each sensor publish its detections in an open, documented form that another vendor's tracker can consume. Ask that the fusion layer fail closed on unverified reports and show disputes to the operator. Ask that delegated authority survive the loss of the link to higher headquarters, and that it be tested with the link pulled.
Ask for track sharing across neighboring installations and with civil partners, with the release rules written down. And ask for a mission record that a review board can read the next morning without a vendor engineer in the room. None of these requirements is exotic, and all of them are cheaper to write into the first contract than to retrofit after the third.
The warning has to reach someone who can act
The Opana operators did everything right. The radar gave forty minutes of warning, and the system around it turned that warning into nothing. Domestic Shield is fixing the sensor half of that problem now, with real money and real urgency.
The other half is the layer that turns detections into one picture, puts authority where the defenders are, and keeps working when the network is attacked along with the base. Installations that buy that layer with their sensors will get the full value of both. Installations that buy sensors alone will have better radars and the same Sunday morning problem.