Intelligence fusion is the practice of combining many sources into one coherent, decision-ready picture. Not a denser map - a clearer one. The discipline spans a whole stack, from raw signal to recommended action, and a whole breadth, from radar tracks to financial identity to weather to narrative. Done well, it turns a flood of fragmented, overlapping, sometimes contradictory inputs into the single thing an operator actually needs: an understanding of what is out there, what it means, and what to do about it.
The word "fusion" gets used loosely, so it helps to separate the layers it actually contains. Data fusion combines low-level data - signals, pixels, detections. Sensor fusion combines the outputs of multiple sensors into tracks and identities. Information fusion combines processed information across sources. Intelligence fusion - the broadest term - combines all of it, across disciplines and domains, into assessed intelligence a commander or an operator can act on. Sensor fusion is one floor of this building. This page is about the whole building.
For the military, this is the engine under Joint All-Domain Command and Control. For a wildfire incident commander, a port authority, a search-and-rescue coordinator, or an insurer pricing real-world risk, it is the same engine under a different name - and most of them never say "JADC2" at all. Both audiences have the identical underlying problem: too many sources, no single picture, and decisions that cannot wait.
The Fusion Stack: From Signal to Decision
The recognized framework for fusion of all kinds is the JDL Data Fusion Model, originated by the U.S. Joint Directors of Laboratories Data Fusion Group in the mid-1980s, formalized in the 1991 Data Fusion Lexicon, revised by Steinberg, Bowman, and White (1998-1999), revisited by Llinas and colleagues ("Revisiting the JDL Data Fusion Model II," Fusion 2004), and extended by the Data Fusion Information Group (DFIG), which added a level for the human in the loop. It organizes fusion into levels, and the levels are the clearest way to see why "fusion" means far more than tracking.
Level 0 - Source and Data Assessment. Conditioning raw observables: detecting a signal, extracting a feature, parsing a feed. The plumbing that turns a sensor's or a source's output into something the rest of the stack can use.
Level 1 - Object Assessment. Associating those observations into tracks and identities - the same object seen by radar, RF, and a cooperative broadcast resolved into one entity. This is sensor fusion proper, and it is where most tools stop.
Level 2 - Situation Assessment. Relationships among objects. Not just where the tracks are, but what they form - who is with whom, what owns what, how the picture hangs together. This is where a track becomes a situation.
Level 3 - Impact and Threat Assessment. What the situation means and where it is going. Projecting consequences, scoring risk, anticipating the next move. The level that turns awareness into warning.
Level 4 - Process Refinement. The system managing itself - tasking sensors, reallocating collection, escalating posture. Closing the loop so fusion drives the very collection that feeds it. Much of this is machine-to-machine.
Level 5 - User and Cognitive Refinement. The human in the loop - reasoning, judgment, decision. The DFIG added this level explicitly because a computer cannot process the parts of a situation that only a person understands, and because the picture has to be presented to that person in a form they can actually use. This is the fusion that happens in the interface, not the machine.
The model is not a strict sequence - real fusion runs these concurrently and loops among them - but it is the common language for what a fusion system has to do, and it makes one thing obvious: a tool that delivers Level 1 and calls itself a fusion platform has done one-sixth of the job.
Fusion Is Not Just Sensor-to-Shooter
The phrase that dominates the conversation is sensor-to-shooter - getting a detection to an effector fast enough to act, the kill chain compressed. It matters, and it is real, but it is one path through the stack, not the whole of fusion.
Sensor-to-sensor fusion is just as important and gets far less attention: using one sensor to cue another, cross-correlating modalities so the picture improves even when nobody is shooting. A passive RF hit cues a radar; an acoustic detection cues an optic; a financial flag cues a closer look at a vessel. The output is a better picture, not a firing solution.
And much of fusion is not about effects at all. Data fusion and intelligence fusion exist to build understanding - the situation and impact assessment of Levels 2 and 3 - whether or not an effector is ever in the loop. A search-and-rescue coordinator fusing AIS, weather, and last-known-position is doing intelligence fusion with no shooter anywhere in the chain.
There is also a split in where fusion happens. Some of it is machine-to-machine (M2M) - automated association, tasking, and escalation running at machine speed below the level of human attention (Level 4 territory). Some of it is in the interface - the UI/UX that fuses for the human, presenting the picture so a person can reason and decide (Level 5). A serious fusion system does both: it automates what should never wait on a human, and it presents what should never be taken away from one.
The Doctrine: JADC2, CJADC2, JADO, and Multi-Domain Operations
In the U.S. military, intelligence fusion is the substance of the command-and-control modernization effort.
Joint All-Domain Command and Control (JADC2) is the Department of Defense approach to connecting sensors, deciders, and effectors across every warfighting domain - land, maritime, air, space, and cyberspace - into a single decision enterprise. Its operating logic is often stated as Sense, Make Sense, Act: collect across domains, fuse into understanding, and act faster than the adversary. JADC2 is not a platform; it is an architecture and a strategy, now led on the DoD side by the Chief Digital and Artificial Intelligence Office (CDAO).
Combined Joint All-Domain Command and Control (CJADC2) is the same effort with the "Combined" prefix added to emphasize allies and partners. The Deputy Secretary of Defense declared a CJADC2 "minimum viable capability" in February 2024, describing it not as a single system but as a federation of concepts, technologies, policies, tools, and talent. The framing matters: the official vision of all-domain C2 is a federation of interoperable capabilities, not one monolith - which is precisely the seam an edge-deployable, integration-native fusion node fits into.
Joint All-Domain Operations (JADO) is the warfighting concept that JADC2 enables - the actual conduct of operations across all domains. JADC2 is the command and control; JADO is what you do with it.
Multi-Domain Operations (MDO) is the Army's capstone doctrine, elevated to doctrine in FM 3-0, Operations (2022). Each service contributes its own line of effort - the Army through Project Convergence, the Navy through Project Overmatch, the Air Force through the Advanced Battle Management System - and all of them depend on fusion as the connective tissue.
Two terms in this thicket are routinely conflated. Multi-domain means operating across the warfighting domains. Cross-domain carries two distinct meanings: cross-domain effects (creating an effect in one domain from another) and the cross-domain solution (CDS), a security-accredited boundary that moves data between networks of different classification levels. An unclassified, dual-use fusion node sidesteps the CDS problem entirely by staying on one side of the classification boundary.
Multi-INT and All-Source: Fusing More Than Tracks
Doctrine on the intelligence side has its own word for fusion breadth: all-source. All-source intelligence fuses across the collection disciplines - GEOINT, SIGINT, MASINT, HUMINT, OSINT, and increasingly FININT - rather than relying on any single INT. This is the horizontal axis of fusion, orthogonal to the JDL levels.
The INTs: what multi-INT actually combines
The intelligence community organizes collection into disciplines, each with its own sources, formats, and tradecraft. A multi-INT picture draws from several at once.
- GEOINT (Geospatial Intelligence): imagery and geospatial information about features and activity on the earth. The map and what is on it.
- IMINT (Imagery Intelligence): the imagery component of GEOINT - optical, infrared, and radar imagery from overhead and airborne sensors.
- SIGINT (Signals Intelligence): intelligence from intercepted signals, split into COMINT (Communications Intelligence) and ELINT (Electronic Intelligence) for non-communications emitters like radars.
- MASINT (Measurement and Signature Intelligence): technical signatures - acoustic, seismic, radar cross-section, spectral - that identify or characterize a source.
- OSINT (Open-Source Intelligence): publicly available information, from registries and filings to news and social media.
- HUMINT (Human Intelligence): information from human sources.
- FININT (Financial Intelligence): financial flows, ownership, sanctions exposure, and the identity behind a transaction. Increasingly first-class in modern fusion because identity and money resolve a lot of ambiguity a sensor cannot.
Each discipline answers a different question. The answer to "what is this and should I act" usually requires more than one of them - which is the whole premise of multi-INT.
Why crossing disciplines is genuinely hard
Multi-sensor fusion has a clean math core. Multi-INT does not, because the inputs are heterogeneous in every dimension that matters.
Identity is the central problem. The same real-world entity - a vessel, a person, a company, an aircraft - appears under different identifiers in different feeds: a hull number here, a callsign there, a registered owner that is a shell, a beneficial owner three layers up. Tying those records to one entity is entity resolution, and nothing useful happens across disciplines until it is solved.
Latency and confidence do not line up. A radar track updates every second. A sanctions list updates on a regulator's schedule. An analyst note is authoritative but sparse. Fusing them means reconciling a fast, low-confidence stream with a slow, high-confidence one, and being honest about which drove the answer.
Contradiction is normal. Two disciplines will disagree. The job is not to hide the disagreement behind a single icon but to resolve it and keep the provenance, so a human can see what the call rested on and overrule it. A multi-INT system that produces a confident answer with no audit trail is a liability, not a capability.
Classification and releasability vary. Disciplines carry different handling rules. Fusing them in software means the system has to respect what can be combined, shown, and shared - by user and by network - not just what is technically joinable.
A useful test when evaluating a "multi-INT fusion platform": ask whether the system merges sources into one assessed track with provenance, or whether it just shows them side by side. Most do the second and call it the first.
Concretely, fusing more than tracks means correlating a kinetic sensor picture with the identity and financial lens (who owns and controls the platform - entity resolution and threat finance), the geospatial and environmental lens (weather, terrain, and space weather as confidence modifiers on every sensor and route), the narrative and open-source lens, and the space lens. A radar track tells you something is there. Fused with ownership, environment, and pattern of life, it tells you what it is, whose it is, and whether to care - which is the difference between data and intelligence.
The Civilian and Dual-Use Angle
The fusion problem is not military. It is universal, and most of the people who have it have never used the word JADC2 in their lives.
A wildfire incident command post fusing aircraft positions, crew locations, weather, and terrain is doing all-domain fusion. A search-and-rescue coordinator correlating AIS, radar, drift models, and last-known-position is doing situation and impact assessment. A port authority watching vessels, cargo manifests, and sanctions exposure is doing multi-INT fusion. A law-enforcement fusion center, a stadium security operation, a critical-infrastructure operations center, an insurer pricing maritime risk - all of them are running some subset of the JDL stack, and none of them call it that.
The civilian situational-awareness and tactical-mapping tools these operators already use do real and valuable work, but they describe themselves as maps and trackers, not as command-and-control architectures. That is not a gap in the doctrine; it is a translation problem. A platform that can speak both dialects - rigorous enough for the doctrinal buyer, plain enough for the operator who just needs the picture - serves a market that the classified, defense-only tools structurally cannot reach.
What Intelligence-Fusion Software Actually Requires
Pulling the stack and the breadth together, a real intelligence-fusion capability has to do all of the following, not a slice of it.
The whole stack, not just Level 1. Object and track fusion is the floor. Situation assessment, impact and threat projection, process refinement, and human decision support are the rest of the building - and most "fusion" tools never leave the ground floor.
All-source, not sensor-only. Fusion across the INT disciplines and the non-sensor lenses - identity and finance, environment, narrative, space - not just across radars.
Both M2M and the interface. Machine-speed automation for what cannot wait, and a cognitive interface for what must stay with the human, each engineered on purpose.
Sensor-to-sensor and sensor-to-shooter. Cross-cueing that improves the picture, and a compressed chain to action when action is required - the platform should not force a choice between understanding and effect.
Edge-deployable and unclassified by design. The full stack running at the lowest level, air-gap capable, without dragging the operator into a cross-domain-solution accreditation project.
Simulation-backed. The same fusion engine exercisable in a wargame, so the decision chain is proven before contact rather than during it.
A federation-friendly posture. All-domain C2 is officially a federation of interoperable capabilities, not a monolith - so a fusion node has to be integration-native, speak open formats, and play well with what the operator already runs.
Glossary
- ABMS (Advanced Battle Management System): The U.S. Air Force contribution to all-domain command and control.
- All-Source Intelligence: Intelligence fused across the collection disciplines (GEOINT, SIGINT, MASINT, HUMINT, OSINT, FININT) rather than from a single INT.
- CDAO (Chief Digital and Artificial Intelligence Office): The DoD office leading JADC2/CJADC2 implementation.
- CJADC2 (Combined Joint All-Domain Command and Control): JADC2 with explicit emphasis on allies and partners; a minimum viable capability was declared in February 2024.
- Cross-Domain Effect: An effect created in one warfighting domain by action in another.
- Cross-Domain Solution (CDS): A security-accredited boundary that moves data between networks of different classification levels.
- Data Fusion: Combining low-level data such as signals, pixels, and detections.
- DFIG (Data Fusion Information Group): The body that extended the JDL model, adding the user-refinement level.
- Information Fusion: Combining processed information across sources.
- Intelligence Fusion: The broadest fusion term - combining all sources and disciplines into assessed, decision-ready intelligence.
- JADC2 (Joint All-Domain Command and Control): The DoD approach to connecting sensors, deciders, and effectors across all domains.
- JADO (Joint All-Domain Operations): The warfighting concept that JADC2 enables.
- JDL Data Fusion Model: The recognized framework dividing fusion into levels from source assessment to user refinement.
- M2M (Machine-to-Machine): Fusion and tasking performed automatically at machine speed without human intervention.
- MDO (Multi-Domain Operations): The U.S. Army capstone doctrine (FM 3-0, 2022) for operations across all domains.
- Multi-Domain: Operating across the warfighting domains (land, maritime, air, space, cyberspace).
- Multi-INT: Fusion across multiple intelligence collection disciplines.
- Sense, Make Sense, Act: The operating logic of JADC2 - collect, fuse into understanding, and act.
- Sensor Fusion: Combining the outputs of multiple sensors into tracks and identities (JDL Level 1).
- Sensor-to-Sensor: Using one sensor to cue or correlate with another to improve the picture.
- Sensor-to-Shooter: Connecting a detection to an effector to compress the time to action.
- Situation Assessment: Fusion of the relationships among objects into a coherent situation (JDL Level 2).
Related: What is Sensor Fusion? | What is JADC2? | What is a Common Operational Picture? | What is Maritime Intelligence? | Fusion & Decision Engine capability | What is Counter Threat Finance? | What is Entity Resolution?
