An on-site assessment run to a formal protocol: nine phases, thirteen threat classes, measured values instead of opinions. You get a prioritised, costed plan that works as a procurement specification.
Measure first, buy second.
The most expensive way to learn about a blind spot is to learn about it during an incident. An audit looks at your site through the eyes of a hostile operator: where he comes in from, what he can achieve, how much time your shift has and what they are actually permitted to do. The result is not a consultant's opinion but measured values you can defend in front of a board and present to a regulator.
Six areas, none of them skipped.
The priority of the areas changes with the type of site, but their composition does not. If there is nothing to report in an area, the report says so explicitly: checked, no significant findings.
A site is not checked «in general» but against a closed list of thirteen threat classes. The list is reviewed at least twice a year.
| Threat group | What it covers | What answers it |
|---|---|---|
| Control links you cannot see | Fibre optic, cellular-network control, autonomous flight in radio silence, satellite link | The RF line is empty against these. Acoustics, radar and optics do the work |
| Conventional radio control | Off-the-shelf aircraft on known bands, including Remote ID | Passive RF line, 5–10 km — the earliest warning available |
| Conditions that make observation hard | Small radar cross-section, night, fog, rain, swarm employment | Radar and acoustics, optics with a thermal channel, rehearsed procedures |
| Physical effect | Payload release, contact strike, throw-over across the perimeter | Close-in detection layer plus procedures for sheltering staff and stopping processes |
| Surveillance and preparation | Reconnaissance of site routine, filming of processes, repeat overflights | Event logging and correlation with previous overflights |
| Friendly and legitimate traffic | Contractor drones, the insider, navigation spoofing of your own aircraft | Register of authorised flights, clearance procedure, monitoring of the navigation picture |
| Attack on the protection itself | Jamming, power cuts, damage to nodes, interference with communication links | Node autonomy, backup links, physical protection of masts and cabinets |
Nine phases in strict order. A phase can be skipped only with written consent, and the omission is recorded in the report.
| Phase | What happens | Indicative |
|---|---|---|
| 0. NDA and preparation | Work begins with signing a non-disclosure agreement — before that we request no information about the site at all. Then: client questionnaire, open geospatial data, airspace picture and flight restrictions. The team arrives on site with working hypotheses about approach corridors, not with an empty notebook. | 1–2 days, remote |
| 1. Kick-off meeting | Signing the data protocol with the storage option chosen, clearances, confirmation of the critical asset list with senior management, agreement on the response test. | 2–3 hours |
| 2. Survey and digital twin | Photogrammetry of the site, or a model assembled from open data. Buildings and their heights, terrain, fence lines, vegetation and the positions of existing cameras go into the model. | 0.5–1 day |
| 3. Field measurements | Spectrum and GNSS, acoustic background at candidate node positions, sensor inventory, test flights of an agreed aircraft, verification of corridors on the ground. | 1–2 days |
| 4. Analysis on the twin | Calculation of actual coverage volumes, blind-spot map, run-through of approach profiles for every applicable threat scenario. | 1–2 days |
| 5. People and procedures check | A «cold start» exercise: the time from a target appearing to a person with authority being ready to decide is measured. | 3–4 hours |
| 6. Legal review | What is lawful at this site, who is authorised to apply countermeasures, the escalation procedure and the evidentiary requirements. | 0.5 day |
| 7. Report | Assembly of the five-document package and an oral defence of the findings. A report nobody has walked through out loud stays unread. | 2–3 days |
| 8. Handover and data mode | Execution of the chosen storage option (A1, A2, A3 or B), handover of the encryption key to the client, signed record, scheduling of the planned review in 12 months. | 0.5 day |
The audit counts as complete when every line has either a measured value or a justified «not applicable».
| Indicator | How it is obtained | When it is a problem |
|---|---|---|
| Time to decision | Stopwatch exercise: from target appearance to a person with authority being ready | Over three minutes, or never measured at all |
| Share of threats covered | Threat-versus-layer matrix against the actual state of the site | Fewer than ten of the thirteen classes covered |
| Layers independent of the radio link | At least one line that works when the target is radio-silent | No such layer exists — a critical finding |
| Approach corridor coverage | Calculation on the digital twin for altitudes up to 50 m | Under 80 % of identified corridors covered |
| Observability | Who actually looks at the screen, and when | The alarm arrives where nobody is at that hour |
| Evidentiary trail | Completeness of the logs after a test event | No tracks, no video or no timestamps |
| Node autonomy | Operation without external power and communications | Under four hours |
Coverage and blind spots as a calculation, not as an estimate drawn over a satellite image.
The traditional output of an audit is a document. Documents describe blind spots, but you cannot fly through a document. A spatial model of the site changes what an audit is able to show at all: you can see where each sensor's field of view ends, where two fields fail to meet, and what a low approach along the treeline looks like from every camera already installed — before anyone is asked to approve a purchase. The model is deployed in the A3O Mini Rapid environment. Coverage and blind-spot maps included in the report are calculated from it. Attack simulation on the model and the operator training mode with adversarial AI are available to pilot-programme participants and run on the A3O Mini system deployed at the site.
The team arrives with its own self-contained kit and does not connect to your network.
Capture, processing and storage all run on autonomous equipment we bring with us: a laptop with full-disk encryption, a hardware key, an isolated field switch with no internet access, and media in a sealed case. At no stage is there any connection to your corporate network, Wi-Fi or cloud services. Site material exists only inside that kit until you decide what happens to it next.
A spatial model of a site is, in effect, a detailed map of its vulnerabilities. The option is chosen before work starts and fixed in a protocol signed together with the contract. In every option where the model is retained, the encryption key belongs to you: we physically cannot open the file without your participation.
The model sits encrypted on our server in an EU jurisdiction. The encryption is enveloped: the file is closed with a session key, and that session key is closed with yours. We hold the container, you hold the key; without it the contents cannot be read even with full access to the disk. Every access is logged, and after 12 months the model is updated on schedule without a repeat survey.
Learn more →The encrypted container stays with you: on your server, inside your perimeter, under your backup rules. Neither a copy nor a key remains with us. For the scheduled update a year later you grant us access to the container for the duration of the work — and revoke it immediately afterwards.
Learn more →The model is written to two encrypted drives and handed over to you against a signed record; there is no network storage at all. This is the option for sites where any presence of data on a network counts as unacceptable risk. The update a year later runs on the drive you issue to the team for the duration of the work.
Learn more →All material is destroyed within 10 working days of the report being accepted, backups included, with a signed certificate. You keep the report, the maps and the calculations; we keep only the fact that the work was performed. A repeat assessment will require the full cycle of work, including a new survey.
Learn more →Five documents — enough to approve or reject a budget without further study.
Scenarios realistic for your site with likelihood and consequence ratings, and a findings register with risk scores.
Learn more →What the existing sensors actually see, where the seams run and which of them open onto approach corridors.
Learn more →Positions and rationale, calculated from assets and consequences rather than from the length of the perimeter.
Learn more →What is lawful at your site, who is authorised, what the escalation procedure is and what to record so an incident can be handed to law enforcement.
Learn more →A phased plan in which the first phase has standalone value and can be implemented without the rest.
Learn more →A site of medium complexity: perimeter up to 4 km, up to 40 existing cameras, one duty shift.
The full cycle takes seven to eleven working days, of which the team spends two to three on site. A major airport or an extended linear asset is scoped separately. Sites of the same type are usually assessed on a representative sample, with the conclusions then extended to the rest.
A demonstration shows what a product can do under conditions the vendor selected. An audit establishes what your site needs, independently of any product, and produces the specification the product then has to meet. That order is precisely what protects you from buying a sensor that will not see the approach that actually matters.
No — and an audit that assumes otherwise is a bad audit. A frequent and entirely legitimate outcome is that the existing cameras and procedures cover more than expected, and the correct next step turns out to be narrow: one sensor in the blind corner, or a change to the list of people who get called at night.
The full cycle is seven to eleven working days for a site of medium complexity, of which two to three are field work. Spectrum capture and coverage modelling run in parallel; the report and roadmap are prepared afterwards.
You decide that before work starts, and there are four options. The model can be stored encrypted on our server in the EU, on your own premises, offline on drives in your safe — or not stored at all: all material is destroyed within ten working days of the report being accepted, with a signed certificate. In all three storage options the encryption key belongs to you, so we are technically unable to open the model without your participation. Your site material is never passed to third parties under any circumstances and is not used to train models without separate written consent.
The Critical Entities Resilience Directive requires a review at least every four years — that is the floor, not the target. In practice the trigger is a change in the picture: a new class of approach spreads, a building goes up nearby, a process is relocated, the duty roster changes. If the model of the site has been retained, the repeat check runs on it and costs substantially less than the first audit.
No, and that is fundamental. Four of the thirteen threat classes — fibre optic, cellular-network control, autonomous flight in radio silence and the satellite link — are not detected by an RF line at all. If a site has no layer that works when the target is completely silent, that is recorded as a critical finding regardless of how much money has gone into RF equipment.
No. Measurements are taken with the plant running as usual — indeed, some of them only make sense that way: the acoustic background is measured both during a working shift and in quiet conditions, because the real range of the acoustic layer depends on it.
The roadmap is built so that the first phase has standalone value and can be implemented without the rest. Many clients start with a pilot deployment at the most problematic sector — which also lets them verify the audit's calculations in practice.
Find the gap before someone finds it for you.
An audit turns «we need counter-drone protection» into a plan with numbers, and becomes the document that justifies everything that follows — both to your own management and to the regulator. Describe your site in the form below and we will come back with a work plan and timeline.
Or write to us directly: audit@dronedefender.eu
Start with a short consultation, live demo or structured pilot program.
