LISA 26 OSINT/HUMINT INTEGRATION

Three Intelligence Sources — Fused Through Dempster-Shafer

Drone sensors provide the overhead view: thermal signatures reveal engine activity, visual cameras enable vehicle classification, and radar tracks show movement patterns. But drones cannot interpret what they see — a column of trucks might be logistics, refugees, or a deception operation. Ground-level intelligence from soldiers, local contacts, and signals intercepts provides the context that gives drone imagery meaning.

OSINT (Open Source Intelligence) adds a third dimension: satellite imagery services, social media geolocation, public shipping databases, radio frequency databases, and government publications. An S2 analyst who identifies a social media post showing enemy vehicles at a specific intersection can cross-reference this with Fischer 26 thermal imagery of the same intersection taken 30 minutes later. If both sources independently confirm vehicle presence, Dempster-Shafer fusion raises confidence from 70 percent (single source) to 91 percent (two independent sources). Three independent confirmations reach 97 percent.

Manual Entry Workflow in Lisa 26

HUMINT and OSINT data enters Lisa 26 through the S2 terminal at brigade or battalion headquarters. The analyst creates an entry with: source type (HUMINT/OSINT), confidence assessment (0-100 percent based on källtillförlitlighet and informationsriktighet using Swedish Armed Forces STANAG 2022 / AJP-2.1 A–F/1–6 grading), MGRS position (if available), free-text description, and classification level. The entry is tagged with a timestamp and the analyst's identity for accountability.

Lisa 26 automatically attempts to correlate the manual entry with existing drone detections within 200 meters and 30 minutes. If a drone detected a vehicle at PA 2345 6789 at 14:00 with 72 percent confidence, and a HUMINT source reported enemy armor at the same intersection at 13:45 with 60 percent confidence, the fused assessment reaches 89 percent. This correlation happens automatically — the analyst enters the data, and Lisa 26 enhances it with everything the drone network already knows. The result appears on the COP as an enriched detection marker showing both the drone observation and the human intelligence source.

Quality Control and Source Evaluation

Not all intelligence sources are equal. Lisa 26 implements Swedish Armed Forcess informationsvärdering per STANAG 2022 / AJP-2.1: källtillförlitlighet (A=helt tillförlitlig through F=tillförlitlighet kan ej bedömas) crossed with informationsriktighet (1=bekräftad through 6=sanning kan ej bedömas). An A1 assessment (reliable source, confirmed information) carries maximum weight in Dempster-Shafer fusion. An F6 assessment (unknown source, unconfirmed information) is displayed on the COP but weighted near zero in fusion calculations. The S2 analyst assigns these ratings at the moment of entry — and Lisa 26 logs the rating for post-mission review of whether the analyst's assessments were accurate over time.

Open the interactive Threat Fusion Dashboard →

Integration Process

OSINT and HUMINT do not replace drone ISR — they complement it. A drone sees what is there now. OSINT can reveal what was there yesterday (satellite imagery) or what the enemy is discussing (intercepted communications). HUMINT provides context that no sensor can capture: morale, intentions, command structure. Lisa 26 fuses all three into one operational picture.

OSINT Tools (Free/Open-Source)

Related Chapters

Implementation

# OSINT/HUMINT Ingestion — Manual Entry to Lisa 26 COP
import json, time

def ingest_humint_report(report, s2_operator):
    """S2 manually enters HUMINT/OSINT into Lisa 26."""
    entry = {
        "source_type": report["type"],  # "HUMINT" or "OSINT"
        "confidence": report.get("confidence", 0.5),
        "timestamp": time.time(),
        "entered_by": s2_operator,
        "position": report.get("mgrs"),
        "description": report.get("text"),
        "classification": "UNCLASSIFIED",  # JEF filter respects this
    }
    
    # Fuse with drone detections at same location
    nearby = lisa26_db.query(
        "SELECT * FROM detections WHERE ST_Distance(geom, %s) < 200",
        [entry["position"]]
    )
    
    if nearby:
        # Dempster-Shafer: combine HUMINT + drone observation
        fused = dempster_shafer_fuse([entry["confidence"]] + 
                [d["confidence"] for d in nearby])
        entry["fused_confidence"] = fused
    
    lisa26_db.insert("intelligence", entry)
    broadcast_cop_update(entry)

Worked Example — Three-Source Fusion on a Real Scenario

Consider an inbound report from a local Ukrainian unit: "tank column visible from hilltop Bravo, approximately 8 tanks, heading north". STANAG 2022 grades this as B3 (usually reliable source, possibly true information). A Fischer 26 orbiting 5 km away sees 8 thermal hotspots on the same road segment, with the AI classifier assigning 72% confidence to "T-72 signature". A Sentinel-2 pass from 45 minutes earlier shows vehicle tracks matching the reported route. Combining these three independent sources via Dempster-Shafer:

sources = [
    {"type": "HUMINT", "grade": "B3", "confidence_equiv": 0.72},  # from STANAG 2022 table
    {"type": "EO_IR",  "grade": "A2", "confidence_equiv": 0.82},  # drone thermal
    {"type": "OSINT",  "grade": "A3", "confidence_equiv": 0.67},  # Sentinel-2 historical
]

# Dempster-Shafer combination rule for independent sources
m_combined = 1 - (1-0.72) * (1-0.82) * (1-0.67)
# = 1 - 0.28 × 0.18 × 0.33
# = 1 - 0.0166
# = 0.983

print(f"Fused confidence: {m_combined:.3f}")
# Fused confidence: 0.983  — sufficient to escalate from L1 advisory to L2 recommendation

The three-source combination lifts confidence from any individual source's 67–82% to 98.3%, which crosses the L2 recommendation threshold. The operator sees a single fused track on the COP with a 98.3% confidence badge and a citation chain showing each contributing source. If any one source is later repudiated (e.g. the local unit misidentified the vehicles), the combination can be re-run without that source and the fused confidence drops accordingly — the audit trail survives.

Why It Matters

A single drone operating alone can achieve at most 70–80% classification confidence on a moving target in mixed terrain. No single sensor can do better because occlusion, thermal background, and classifier bias cap the achievable accuracy. OSINT and HUMINT fuse at the decision layer to raise the combined confidence above the threshold required for weapons release authorisation. Without this fusion, FSG-A operators would be forced either to lower the confidence threshold (increasing civilian harm risk) or to refuse engagements that are actually justified. The intelligence-evaluation framework from STANAG 2022 provides the grammar for comparing dissimilar sources; Dempster-Shafer provides the arithmetic. Together, they let a small force punch above its weight by integrating sources that a larger force would keep in separate stovepipes.

Sources

Mathematical proofs. Dempster-Shafer fusion (70% single source, 91% two independent, 97% three independent) is verified in provable_claims.py under DS_FUSION_2 and DS_FUSION_3. Formula: m = 1 − ∏(1 − m_i) for independent sensor observations.

Standard sources. STANAG 2022 / AJP-2.1 intelligence evaluation — A–F source reliability and 1–6 information credibility grading. Sentinel-2 10 m resolution — ESA technical specification. RTL-SDR at €25 — commercial market price.

Operational estimates — not validated by field testing. The 200-metre / 30-minute correlation window is an FSG-A design choice, not calibrated against operational data. It may require revision for dense battlespace where multiple targets could transit the same point within 30 minutes. The OSINT/HUMINT module is conceptual architecture: FSG-A has not tested the system in real operations.

External standards and references. ArduPilot documentation. ExpressLRS documentation. NATO STANAG 4609 Ed. 4 (motion imagery metadata), STANAG 4671 (UAV airworthiness), and STANAG 2022 (intelligence source reliability). Specifically: Watling & Reynolds, "Meatgrinder: Russian Tactics", RUSI (2023); Bronk, Reynolds & Watling, "The Russian Air War and Ukrainian Requirements for Air Defence", RUSI (2022); ISW daily campaign assessments (understandingwar.org archive); CSIS Center for Strategic and International Studies Ukraine briefings. — FSG-A has no operational experience of its own. Bellingcat OSINT methodology.