Identity-blind · Critical-infrastructure protection
Attacks are rehearsed.
We catch the rehearsal.
Attacks on critical infrastructure are preceded by reconnaissance from concealed positions with sightlines to what matters. wAItchdog computes those positions, is built to deploy its own sensors to close the coverage gaps, and tells reconnaissance from noise — identity-blind, and lawful by design.
To see a more detailed, narrated walkthrough of how it works — the geometry, the sensing, and identity-blind detection on a representative site — watch the interactive demo →
Securing the perimeter
The problem
An attacker's best vantage point is the one you're not watching.
Before an attack, hostile actors study a site from concealed positions with clear lines to its critical components — the blind spots fixed cameras don't cover. Motion-anomaly tools drown operators in false alarms; identity-based surveillance can't be lawfully fielded. Detection is hardest exactly where it matters most.
U.S. substations have been physically attacked — Metcalf (2013) and Moore County (2022) — degrading the grid that installations and communities depend on. Such attacks are preceded by observable reconnaissance.
Coverage completion
Find the gaps. Close them. Tell signal from noise.
You can't watch a mile of perimeter. You can watch the few positions that matter — and cover them with sensing built to see what cameras miss.
Prescribe
A geometry engine ranks the concealed positions with sightlines to your critical components — an attacker's real vantage points.
Cover
We're built to deploy our own best-fit sensors — radar-flagship, all-weather, low-power — to exactly those gaps.
Discriminate
A behavioral engine separates reconnaissance from wildlife, weather, and passersby. Precision through corroboration — identity-blind.
Refer
Confirmed reconnaissance goes to lawful authority across a documented boundary — no in-house identification, no pursuit. A physical attack triggers the separate, attack-gated response.
Technical foundations
Three engines, one defensible core.
The sensing is commodity and the interoperability is standard. The value — and the patents — is in the analytics: computing where an adversary would watch from, covering exactly those positions, and separating reconnaissance from noise without ever resolving identity.
Observation-position engine
From a georeferenced site model and designated critical components, it computes the concealed standoff positions an adversary would actually use — by viewshed and line-of-sight analysis against terrain, structures, and occluders.
- Viewshed / occlusion modeling
- Visual access ∩ coverage deficit ∩ concealment
- Ranked by attacker utility
Coverage-completion optimizer
Given the ranked positions, it solves for the minimal set of sensors and modalities whose combined field of regard spans them — a set-cover / facility-location problem under line-of-sight and siting constraints.
- Minimal-sensor set cover
- Per-gap modality selection
- Radar-flagship, modality-agnostic
Identity-blind classifier
On kinematic tracks alone, it scores behavior directed at a critical component against a learned model of normal activity, and withholds an alert until independent corroboration — deliberately tuned for specificity over sensitivity.
- Kinematic behavioral features
- Benign-null / pattern-of-life
- Corroboration-gated alerts
Together, the three answer three questions — where an adversary would watch from, the fewest sensors needed to cover it, and whether motion there is surveillance or noise — from geometry and behavior alone, never identity. The models and thresholds that make it perform are held as trade secrets, and the method is patent-pending.
System architecture
Built on proven sensing, not promised hardware.
wAItchdog is the autonomy and analytics layer, not another box to certify. It ingests standard tracks from fielded commercial sensors over defense-standard interfaces, registers them into a common site frame, and runs the geometry and discrimination engines at the edge — nothing to rip out, no proprietary radar to invent.
The flagship modality
Sensors that see where cameras can't.
Works through darkness, fog, and foliage
Radar leads because it sees what optics can't — and it's sited to cover the specific gap.
Reports motion, not identity
Returns are position and velocity — never faces or plates. The added sensing strengthens the privacy posture.
Deploys anywhere
Solar- or battery-powered nodes at remote positions — no grid connection, no trenching.
Modality-agnostic
The brain prescribes the sensor.
Radar is the flagship, not the whole story. The geometry engine prescribes the right modality per gap — radar, RF-presence, acoustic, or a cued camera — and standardizes on open sensor feeds. Incumbents can buy the same hardware we do; what they don't have is the targeting and the reconnaissance-vs-noise discrimination.
The moat is the analytics — not the radar.
If the rehearsal is missed · roadmap
And if an attack comes — a graduated, attack-gated response.
Detection can't catch every attack. So beyond steady-state detection, the platform escalates — but only on a physically-confirmed attack, and within hard limits. Surveillance is ambiguous; an attack is a crime in progress.
Identity-blind detection — the engines above. No personal data, ever.
On a confirmed attack, preserve the tracks already held and cue law enforcement in real time — where the attackers are and which way they're heading. Never who.
A drone launches automatically, follows the attackers as they flee, and streams live location — plus plate, vehicle type and color — to responders. A human stays on the loop; lawful authority directs the outcome.
Attack-gated · automatic launch · human on the loop · lawful-authority-directed · fully audited. A roadmap capability whose method is now captured in its own filed provisional, distinct from the detection core — and a natural fit for well-funded DoD counter-threat and autonomy programs. See it in the demo →
Intellectual property
The method is filed. The internals stay secret.
Two U.S. provisional patents now cover the platform. The first covers the detection pipeline end to end — geometry-derived observation-position targeting, minimal-coverage sensor placement, and identity-blind corroborated discrimination. The second covers the attack-gated graduated response and further method improvements — both written to protect the approach while the models and thresholds that make it perform remain unpublished.
Detection method claim (first provisional)
Covers the geometry-to-coverage-to-discrimination sequence and the lawful-referral boundary — not one narrow implementation.
Response & further claims (second provisional)
The attack-gated graduated escalation, plus additional discrimination, site-modeling, and resource-allocation methods.
Trade-secret model internals
Scoring, weights, and thresholds are deliberately kept out of the disclosure.
Patent-pending
Two U.S. provisionals filed
The detection core — observation-position targeting, coverage-completion placement, identity-blind discrimination, corroboration, and the refer-don't-pursue boundary — is captured in the first provisional. A second provisional covers the attack-gated graduated response and its supporting method claims.
Broad claims · trade-secret internals
Dual-use
One platform. Two markets that need it.
Installation mission assurance
Protecting the infrastructure a defended mission depends on — the fixed assets and the commercial grid an installation can't fully control.
- SBIR / STTR pathway
- On-installation demonstrations
- CUI / CMMC · FOCI-clean
Critical-infrastructure protection
Utilities, water, pipelines, telecom, and data centers — the high-consequence sites first, expanded one site at a time.
- Value-first coverage-gap assessment
- Deployed sensing + managed monitoring
- Lawful referral, not surveillance
Responsible by design
Protection without becoming a surveillance tool.
The capability is identity-blind and lawful by construction — a public-trust asset, not a liability. Even the sensors are behavior-by-construction.
Identity-blind
Sensors report motion and position, never who a person is.
Lawful & on-property
Runs on lawfully-held, on-site sensing with a defensible posture.
Attack-gated escalation
In steady state, leads go to lawful authority — no in-house identification, no pursuit. Only a physically-confirmed attack triggers the graduated response.
Minimized & audited
Enforced data minimization and immutable, oversight-reviewable audit.
CUI / CMMC
Aligned to NIST SP 800-171, with a defined authorization boundary.
FOCI-clean
No foreign ownership, control, or influence — clearance-eligible.
Where it stands
Built and tested — and specific about what's next.
No inflated pilot counts, no synthetic detection rates. This is exactly what exists today, and what we're partnering and raising to field.
Engines built and tested
The geometry, coverage-completion, and identity-blind discrimination engines are built and running; the method is captured in two patent-pending provisionals.
Standards-based integration path
Designed to ingest Cursor-on-Target / TAK from fielded commercial radar — no hardware to invent.
Seeking design partners
Infrastructure operators and installations for a scoped, on-property pilot.
Raising to deploy & validate
Pre-seed capital plus non-dilutive SBIR / STTR to stand the engines up on a live site.
Get in touch
Request a briefing.
For infrastructure operators, program partners, prospective advisors, and investors. We'll walk you through the geometry, the sensing, and how a coverage-gap assessment maps your blind spots — no sensitive access required.
Colorado, USA · Confidential positioning · figures illustrative unless noted