Defconder
Provable, not just claimed. Sovereign, not dependent.
Defconder connects sensors, specialist systems and reports into one shared operational picture. Event sourced, formally verifiable and entirely on your own infrastructure, with AI under your control and rules that are proven, not claimed. It includes the 3D world Terra with a crisis plan, an open verifier that recomputes the log without Defconder, the command communication Comms, case questions with quote duty, an AI layer with memory per installation, a data engine for millions of records, a workflow studio with version spaces and a manifesto that explains why Defconder builds no weapons. The sovereign European alternative in the market for situational awareness and ontology systems.
Go to the domains
Defconder on the web
The dedicated domains (defconder.de in German, defconder.com in English) show the platform in depth: architecture, modules, sovereignty and security, industries and contact.
The pricing model is not fixed yet. Pilot partners discuss it with us before it is set, and help decide what comes next.
Building blocks
What makes Defconder different
Core
Ontology Builder
Build industry specific data models visually, link them and secure them with self diagnosis, for agencies, hospitals, industry or security authorities.
Foundation
Event Sourcing
Every change is an immutable event hash chained with SHA-256. State can be reconstructed from the chain at any time, it is never merely asserted.
Verification
AxiomGate and Z3
Classification consistency is not only recommended on the client, it is formally proven by an SMT solver before any security relevant access.
Role system
Clearance levels
Fixed release levels instead of generic RBAC, checked on the server and formally. Sign in with password and TOTP or with phishing resistant passkeys.
Integration
Sensor Bridge
External systems create records automatically via API key, checked with the same clearance logic as any human access.
Layout
Layout and app builder
Build your own forms and dashboards with drag and drop and assemble several applications from a single ontology, without programming skills.
Command
Defconder Comms
End-to-end encrypted command communication on your own infrastructure: orders with a signed acknowledgement, alerts with an escalation chain, mesh from phone to phone, a decoy PIN. In development, without independent assessment.
3D world
Defconder Terra
Terrain, buildings, weather, water, fire and supply outage from official data, and from them a crisis plan with a map, checked by 29 fixed rules without AI. Pilot area Hanover Mitte.
Recompute
Open verifier
Recomputes the log without Defconder, in JavaScript, Python and Rust, with no network. With witnesses, time proof via Bitcoin, sampling and a hash-based signature.
Modules
One platform, many capabilities
Dedicated modules have grown out of the ontology core. They run on exercise data and real public data sources and share the same log, the same clearance levels and the same chain of evidence.
Operational picture
Real time picture under APP-6
Every new record goes out over WebSocket to every authorised user instantly. Objects appear on the map under NATO APP-6, charts and tables show the same data with the same permissions.
Sources
Information from every source
Nine ways in, formats are detected automatically, millions of entries stay searchable, every entry is bound to the log. Details further down.
Space
Defconder Orbit
Satellites and debris are continuously recomputed from public orbital data (SGP4, CelesTrak). When an at-risk object approaches a site, an event is created in the log automatically.
Airspace
Defconder ASDA
Aircraft from public ADS-B data of the OpenSky Network. When one drops below the safety distance, a mathematically verified event is created, with its own SHA-512 proof chain.
Fusion
Sensor fusion
Two reports about the same object are merged into one track, with the positional disagreement reported openly. Checkable math instead of a trained black box.
Search
Federated search with purpose binding
One search across all domain packs. Every hit is checked individually against clearance level and purpose before it comes back, never filtered afterwards.
3D world
Defconder Terra and crisis plan
Terrain, buildings, weather, water, fire and supply outage from official data, and from them a crisis plan with a map, checked by 29 fixed rules without AI. Pilot area Hanover Mitte, models simplified and not calibrated.
Recompute
Open verifier
Recomputes the log without Defconder, in JavaScript, Python and Rust, with no network. With witnesses, time proof via Bitcoin, sampling with public randomness and a hash-based signature.
Command
Defconder Comms
End-to-end encrypted command communication on your own infrastructure: orders with a signed acknowledgement, alerts with an escalation chain, mesh from phone to phone, decoy PIN, archive under the four-eyes principle. In development, without independent assessment and accreditation.
Resilient
Mesh network
With no internet, no cellular network and no server, devices reconcile their picture directly over Bluetooth. Every report carries its cryptographically verified origin, forgeries get caught.
AI
AI with a sovereignty gate
The AI suggests whether and why. Formulas decide what happens. From a secrecy level you set, the server blocks cloud models by itself.
Map and media
Own map server and media analysis
Your own map data fully offline with tactical graphics and terrain analysis. Media analysis delivers evidence packages that the open verifier recomputes without a network. Real-time biometric matching is still missing.
Stance
Protective Shield and Klartext
Weapon-free, reversible protection with an advance mandate from two humans and an instant veto (prototype with a proof run), plus signed messages to the other side that contain only verifiable truth. Both follow the manifesto The right to hesitate.
AI
Case questions and AI layer
Questions in full sentences about the case, every statement with a verbatim quote and check value. Memory per installation, a gate before every model call, model change only by two. Local models only.
Data
Data engine
Ingest millions of records, link them into a graph, match persons without merging strangers, assign addresses offline, search by meaning. When in doubt block instead of guess, judge's reservation in the blind graph.
Workflows
Workflow studio and version spaces
Workflows checked before start for actions without approval, with a run record to recompute. Changes to the model through branches and a merge by a second person.
Industry
Hospital scenario
A domain pack from configuration alone with HL7 v2 and FHIR, emergency power runtime and IT emergency operation as a rule. Exercise scenario, not approved.
Evaluation
Process analysis
Process mining on your own log, pseudonymised, with minimum group size and recomputable key-figure evidence. Local and switched off until released.
Principles
Five rules that are not negotiable
- 01
Event sourcing throughout
Nothing is overwritten, not even configuration, ontology definitions or user rights. State is always a projection of the event log.
- 02
Industry agnostic from day one
The core only knows generic building blocks such as entity types, relationship types, attributes and rules. Industries are swappable domain packs, not special cases in the code.
- 03
Provable, not just claimed
Critical logic is checked deterministically, hard constraints run through the Z3 solver and return a proof instead of an estimate.
- 04
Self explaining
Contradictions and risks are detected immediately, explained in plain language and come with a suggestion. No silent failure state.
- 05
Sovereignty as an architecture decision
Own it, run it, prove it. Runs on your own or customer owned infrastructure, fully without an internet connection if needed.
Positioning
How Defconder positions itself
The point of comparison is explicitly not that Defconder would be better. Established platforms have decades of head start. The point is a different one: a European, sovereign alternative that can be operated entirely on site, without dependence on a single foreign vendor, and that proves its security promises formally instead of merely asserting them.
One core. Every industry.
Domain packs instead of special solutions
The ontology core knows no hard coded industry terms. Industries are swappable domain packs, and the same software already carries several of them in parallel today, without a single line of code difference.
Information from every source
Everything in. What matters up front. Every detail provable.
A situation never emerges from a single data source, but from dozens: specialist systems, sensors, radio, files, network protocols, reports. Defconder takes them in every common format, works out for itself what is in them, and keeps millions of entries searchable. Every single entry is bound to the tamper-evident log. What matters becomes part of the picture: by rule or with one click.
5 million
entries searched and verified in a single stream
over 12,000
reports per second taken in by the running server
under 10 ms
response time by key, time or map section, at five million entries
66
log entries for one million reports
Measured on a single workstation, not on data centre hardware. Agencies and armed forces run powerful servers with fast storage. With more cores and faster storage there is headroom.
From source to picture in four steps
- 1
Connect
Pick a source, enter an address, folder or topic. For well-known services a template is enough.
- 2
Detect
Defconder reads a sample and proposes the format, time, location and key fields as well as pagination.
- 3
Anchor
Every batch gets a check value that is written into the main log. Changes become visible.
- 4
Adopt
A rule or one click turns the entry into a record of the picture, with a reference to its origin.
What the source integration can do
Connect instead of program
Defconder detects format, time field, location field and pagination from a sample by itself. For well-known services a template is enough. A new source is configuration, not a project.
Nine ways in
Polling web interfaces, webhook, drop folder, TCP, UDP, WebSocket, event streams, MQTT for devices and Kafka for data platforms. Credentials stay on the server as secrets.
Millions of entries, found instantly
Full text, time, location, key and any field. The answer comes in milliseconds, even in the largest stream tested.
Every entry provable
Not just the hit, the mass behind it is bound to the log in a tamper-evident way. Altered, deleted or injected: it gets noticed. Every entry has a proof you can save.
What matters up front
A rule or one click turns an entry into a record of the picture, with a reference to stream and origin. A cap per batch keeps the picture clear instead of flooding it.
Clearance level applies here too
Every stream has its own classification. Adopted data is never classified lower. Whoever lacks a level does not even see the stream in the list.
Hits are not enough. Decisions need proof.
Pure data platforms deliver hits. In a situation you also need to know whether an entry is genuine, who may see it and what follows from it. Defconder answers all three, on millions of entries and on your own hardware.
Existing analysis systems stay in use. Defconder picks up their hits, anchors them and connects them with everything else in the picture.
What Defconder connects
From the specialist application to the sensor, from ship radio to the social network.
Specialist systems and agencies
Whatever existing systems provide as an interface, export or file is taken in without rebuilding.
- Web interfaces (REST)
- Drop folders
- CSV, XML, JSON
- GZIP files
- Webhooks
Sensors and devices
Measuring stations, cameras with metadata, gateways and devices that report their values.
- MQTT
- TCP and UDP
- WebSocket
- Syslog and CEF
- key=value protocols
Sea, radio and navigation
Position reports from radio and navigation, decoded directly.
- AIS ship reports
- NMEA and GPS
- Event streams (SSE)
Data platforms
Where a data hub already exists, Defconder attaches as a consumer.
- Kafka with groups and read position
- Polling with pagination
- Time windows and identifiers
- Token sign-in
Social networks and messengers
Public posts as a source of the picture. Mastodon and Bluesky accounts without access, Reddit, Bluesky search, X and Telegram with your own access.
- Mastodon
- Bluesky
- X
- Telegram
News and services
Reports, warnings and readings of public services in the same picture as everything else.
- RSS and Atom
- Agency notices
- Weather and water level services
Formats Defconder detects by itself
- JSON
- JSON Lines
- GeoJSON
- CSV and TSV
- XML and RSS
- Syslog and CEF
- NMEA and AIS
- key=value protocols
- Custom line patterns
- GZIP files
The same principle. Three situations.
Manhunt, disaster, defence: different sources, different rules, the same path from mass to decision.
Manhunt: the hit out of millions of reads
Vehicle data, tips, guard logs and answers from specialist systems arrive as streams. Millions of entries stay searchable, down to the second and the map section. Only the hit on a wanted vehicle becomes a record of the picture.
- Sources
- Licence plate reads, tips, specialist systems
- Rule
- Licence plate is on the wanted list
- Picture
- Sighting with place, time and a reference to the original entry
Which sources may be connected is determined by the agency's legal basis, not by the technology.
Disaster: water levels, warnings and reports together
Gauges, weather warnings, dispatch logs and reports from the public come together in one rhythm, by polling, via Kafka or as a file from a network drive. When a water level rises above the alarm level, the incident site appears in the picture. Everything before stays searchable.
- Sources
- Gauges, weather service, dispatch, public reports
- Rule
- Water level above alarm level 3
- Picture
- Incident site with water level history, ready to warn the population
Defence: thousands of contacts, few decisions
Ship radio, flight data, sensors, reconnaissance reports as files and network protocols flow into separate streams with their own classification. Even with millions of entries a day, only the unknown contact in the restricted zone becomes an object of the picture. The mass stays with the operator, on its own hardware, even air-gapped.
- Sources
- AIS, ADS-B, sensors, reports, protocols
- Rule
- Identifier unknown and inside the restricted zone
- Picture
- Contact with source, reliability and classification
Exercise scenarios with fictitious data. No real units or incidents.
Insight
Excerpts from the current development environment.
Rough architecture
[ Domain Pack A ] [ Domain Pack B ] [ Domain Pack C ]
└──────────────────┬──────────────────┘
Ontology definition layer
│
┌───────────────────────┐
│ ONTOLOGY BUILDER │
└────────────┬──────────┘
│
┌───────────────────────┐
│ EVENT LOG │ hash chained, immutable
└────────────┬──────────┘
┌───────────────┴────────────────┐
┌──────────────────────┐ ┌──────────────────────┐
│ VERIFICATION ENGINE │ │ PROJECTIONS & VIEWS │
└──────────────────────┘ └──────────────────────┘Roadmap
First dependable, then broad, then large
A horizon only counts as complete when it holds up in operation at a partner, not when the code is finished. The roadmap describes intentions, not commitments.
Ontology builder with rule checking, Chained event log with independent verifier, Clearance levels on the server, roles, purposes and four-eyes principle, Hardware keys for sign-in, Secrets, main log and system backups encrypted at rest, Actions, AI agents, cloud block from a set secrecy level, AI policy layer with operator lock, learning per installation, Building blocks for any object kind, version spaces, workflows with run record, Recovery in an emergency tried, hardening test in seven stages
Data streams with millions of entries, graph, geocoding, person matching with a block, Semantic search, media analysis with evidence package, Federated search, own map server, time travel and forecasts, Comms (core), Terra, crisis plan with plan coach, Hospital domain pack as a scenario
My dashboard, process analysis, case questions, Help and documentation in the application, Offline maps, Comms radio paths and mesh in device test, Highly available operation
Protective Shield as a product with real sensor hardware and legal review, Roles and rights separated across organisations, Directory service and single sign-on, provenance for data, Independent security review, Biometrics: real-time matching and protection against presentation attacks
Public programming interface, Marketplace for industry models, Partner programme for integrators
Federated operation across organisations, Offline operation in the field, distributed data streams across several servers, Accreditation for classified information, Delivery into air-gapped environments
Build progress
More detailed Defconder articles
Questions and answers
Defconder in short
A sovereign, event sourced situational awareness platform. Sensors, specialist systems and reports come together in one shared real time picture, entirely on your own infrastructure, with a visual ontology builder at its core.
Inquiry
Ask about Defconder
Send me a short note about the environment in which you need situational awareness or a data platform. I answer personally.