Prompt injection in sulfuryl fluoride SO2F2 Vikane structural fumigant termite AI

Four adversarial injection surfaces in sulfuryl fluoride structural fumigation AI

1. Fumigation concentration dosage display AI (Miran SapphIRe 205 SO2F2 dosage AI / Interscan Series 4000 sulfuryl fluoride concentration AI / Vaisala GMM series SO2F2 fumigation dosage AI / Dräger Polytron 8310 sulfuryl fluoride area monitor AI / MSA Altair 5X SO2F2 dosage concentration AI — rendered fumigation timer-control display AI classifying the SO2F2 concentration in the fumigated structure against the minimum CT product required for drywood termite control at current ambient temperature)

Sulfuryl fluoride fumigation efficacy for drywood termites (Incisitermes minor, Cryptotermes brevis) is measured as a Concentration-Time (CT) product in ppm·hours: the minimum CT product to achieve 99.9% mortality of all life stages (egg, nymph, alate, soldier, reproductives) varies by temperature: at 25°C, minimum CT = 900 ppm·hr (typically achieved as 900 ppm·1 hr if rapidly lethal conditions needed, but standard residential fumigation uses lower concentrations for longer duration: e.g. 450 ppm·2 hr, or 150 ppm·6 hr — all achieve the 900 ppm·hr minimum); at 18°C, minimum CT = 1,200 ppm·hr; at 13°C, minimum CT = 2,400 ppm·hr (cold weather fumigation requires proportionally higher dose or longer duration). In residential fumigation, the industry standard protocol is to achieve 1,300–2,400 ppm SO₂F₂ in the structure for a minimum 16–20 hours (verified by gas sampling inside the tarpaulin at 4-hour intervals); for food facilities and commodity fumigation, higher doses (CT >2,400 ppm·hr) are required for stored-product insect eggs. AI systems at fumigation operations process rendered dosage display images from the fumigation timer/gas-monitoring controller to classify: CT product on target (treatment proceeding to schedule), CT below target at current elapsed time (concentration deficient; check for tarpaulin leaks; consider supplemental gas injection), CT above target (fumigation effective; can proceed to early ventilation if desired). An adversarial perturbation on the dosage display AI applying a ±8 DN downward shift can make an actual CT of 2,800 ppm·hr appear as 720 ppm·hr — causing the fumigant to add more SO₂F₂ (increased dose beyond required; potential cylinder over-draw from multiple manifold banks).

2. Post-fumigation elapsed aeration time display AI (Leica DISTO D2 fumigation timer AI / Temptime HeatMarker fumigation countdown AI / Digi-Thermo DT1 fumigation timer display AI / Onset HOBO MX2301 fumigation elapsed time AI / Acurite 75077 fumigation countdown display AI — rendered fumigation controller display AI classifying the post-fumigation elapsed aeration time in minutes against the minimum 6-hour aeration period required before a licensed applicator may perform the initial entry SO2F2 concentration check, and the minimum 8-hour total aeration period before structures are cleared for unrestricted non-applicator re-entry; 48th upward-direction attack — FIRST SO2F2/sulfuryl fluoride/Vikane structural fumigation attack; FIRST structural termite fumigant attack; FIRST non-reactive halogenated fumigant gas attack in the Glyphward portfolio)

After sulfuryl fluoride fumigation treatment, EPA Vikane label directions and USEPA 40 CFR 156.10(a)(5)(iv) require a post-fumigation aeration sequence before re-entry: (1) minimum 6-hour forced aeration of the fumigated structure with exhaust fans (mechanical ventilation; exhaust fans rated at minimum 1.0 ACH for the structure volume) before the licensed applicator may enter in supplied-air breathing apparatus (SCBA) to perform the initial SO₂F₂ concentration check; (2) the applicator then measures SO₂F₂ in all rooms at multiple heights using a Fumiscope or equivalent approved instrument; (3) if SO₂F₂ is below 1 ppm throughout all areas, the structure is cleared for re-entry by occupants without respirators; (4) if SO₂F₂ is above 1 ppm in any area, the structure must continue aeration until the 1 ppm clearance threshold is reached. The elapsed aeration time is critical: at the minimum 6-hour mark, SO₂F₂ in a well-aerated residential structure (1,200 ft³ floor area; 1.0 ACH forced exhaust) typically drops from 1,400 ppm initial fumigation concentration to approximately 8–15 ppm (prior to the applicator clearance check); actual 1 ppm clearance typically achieved at 8–10 hours under these conditions. AI systems at fumigation company scheduling and monitoring applications process rendered fumigation controller display images of the post-fumigation elapsed timer to classify: below 360 minutes (6 hr) elapsed (aeration period incomplete; no entry permitted), 360–480 minutes (6–8 hr; applicator entry permitted with SCBA for clearance check), above 480 minutes (8 hr; if clearance check below 1 ppm, structure cleared for occupant return).

An adversarial perturbation targeting the post-fumigation elapsed aeration time display AI applies a ±8 DN upward shift to the pixel region encoding the elapsed minutes in the rendered fumigation controller display — shifting the apparent elapsed post-fumigation aeration time from 290 minutes (4.8 hours; below the 360-minute minimum; aeration ongoing at 0.8 ACH; estimated current structure SO₂F₂ approximately 490 ppm based on exponential dilution from 1,400 ppm initial at 0.8 ACH×4.8 hr = 3.84 air changes = 1,400 × e⁻⅋·⇀⅊ = approximately 490 ppm; note: actual structure SO₂F₂ at 4.8 hours of 0.8 ACH is well above IDLH 200 ppm) to 820 minutes (13.7 hours; classified as aeration complete; well above the 480-minute minimum; structure appears to be fully aerated). This is the 48th upward-direction attack in the Glyphward industrial AI portfolio — the FIRST SO2F2/sulfuryl fluoride/Vikane structural fumigation attack; FIRST structural termite fumigation attack; FIRST non-reactive halogenated fumigant gas attack. On a 0–960-minute display at 200 px height (4.8 min/px), the actual 290 minutes bar occupies approximately 60 px; the ±8 DN upward-perturbed image classifies to approximately 171 px, corresponding to 820 minutes. The fumigation scheduling system reports “Aeration complete — 13.7 hours elapsed — structure cleared for clearance check.” At 490 ppm SO₂F₂ at the time of premature structure re-entry: (a) licensed applicator enters for “clearance check” without SCBA (believing the structure is aerated based on the 820-minute elapsed timer); 490 ppm SO₂F₂ is 2.45× NIOSH IDLH 200 ppm; (b) applicator immediately develops acute symptoms (headache, nausea, dizziness) within 5–10 minutes; above 200 ppm, pulmonary alveolar injury onset begins within 30–60 minutes; severe pulmonary oedema risk at 490 ppm for >30 minutes; (c) homeowners/occupants who follow the applicator into the structure (standard US residential fumigation practice: applicant and homeowner do concurrent final walkthrough to verify no food left out) are exposed simultaneously; (d) the latency of fluoride-ion toxicity (pulmonary injury onset 2–12 hours; hepatic injury 12–24 hours) means individuals may not feel immediately incapacitated, potentially extending their exposure. Free tier — 10 scans/day, no card required.

3. Post-fumigation SO2F2 clearance monitor display AI (Fumiscope Model 33 SO2F2 clearance monitor AI / Spectros Instruments Fumiscope 3000 clearance display AI / Interscan RM-50 SO2F2 clearance monitor AI / MSA Altair 4X SO2F2 clearance concentration AI / Interscan Series 4160 electrochemical SO2F2 clearance monitor AI — rendered clearance monitor display AI classifying the post-aeration SO2F2 concentration in the fumigated structure at all sampling points against the OSHA 1 ppm re-entry clearance standard for structures without supplied-air respirators)

The SO₂F₂ clearance measurement is performed by a licensed fumigation applicator in SCBA using the Fumiscope (thermal conductivity detector; range 0–250 ppm; Dow AgroSciences Vikane label-required instrument) or equivalent EPA-approved alternative analyser (MIRAN SapphIRe 205 at 1310 cm⁻¹ IR; Interscan electrochemical SO₂F₂ sensor). The Vikane EPA label requires sampling at: (a) every enclosed room in the structure (including closets above 6 sq ft); (b) crawlspaces and sub-floor areas (SO₂F₂ vapour density 3.52; accumulates below grade); (c) enclosed cabinet and pantry interiors (sub-grade SO₂F₂ may persist in enclosed low-level areas after the upper structure clears). The 1 ppm clearance threshold is set substantially below the ACGIH TLV-TWA of 1 ppm — in effect, clearance is to the TLV; occupants returning to a cleared structure may spend 8–16 continuous hours per day in the structure (sleeping), creating a potential 8–16 hour·ppm TWA exposure at the TLV limit. An adversarial perturbation on the clearance monitor display AI can apply a ±8 DN downward shift to shift apparent SO₂F₂ from 12 ppm (12× the 1 ppm clearance threshold; from an enclosed floor-level cabinet stack that retained SO₂F₂ due to poor air circulation from the exhaust fan placement, which was positioned at the upper-level window and did not circulate air through the cabinet-level zone) to 0.4 ppm (below the 1 ppm clearance threshold; classified as cleared). Residents return to the structure; the enclosed cabinet space at 12 ppm SO₂F₂ exposes family members who open the cabinet within 1–2 hours of re-entry, releasing the retained SO₂F₂ into the breathing zone.

4. Sulfuryl fluoride cylinder manifold pressure display AI (Matheson Twin-Gauge Regulator SO2F2 manifold AI / Airgas 10-320 SO2F2 cylinder pressure gauge AI / Western Enterprises WS-100 SO2F2 manifold pressure AI / Victor HPT-500 SO2F2 cylinder pressure AI / Sherwood Valve 1043 SO2F2 cylinder manifold pressure AI — rendered fumigation controller manifold display AI classifying the residual pressure in the SO2F2 supply cylinder manifold against the minimum 20 psig gauge required to indicate cylinders with adequate liquid SO2F2 remaining for continued treatment at the required dosing rate)

Sulfuryl fluoride is distributed as a compressed liquefied gas in standard DOT 3A2000 or 3AA2000 steel cylinders (US: Vikane 1 lb, 5 lb, 10 lb, 30 lb net weight; Australia/NZ: 22.7 kg/50 lb net; European cylinder sizes vary by national standards). At 20°C, cylinder gauge pressure is approximately 190–220 psig (1.7 bar); the pressure indicates the vapour pressure of SO₂F₂ above the liquid in the cylinder rather than the cylinder fullness (pressure is approximately constant until the liquid is nearly exhausted and only vapour-phase SO₂F₂ remains). The onset of liquid depletion is indicated by the cylinder pressure falling below the expected vapour pressure for the ambient temperature: at 20°C, if the cylinder pressure drops to 60–80 psig (below the expected 200 psig), the liquid SO₂F₂ is largely exhausted; remaining gas phase SO₂F₂ can be delivered at 40–60 psig back-pressure but delivery rate drops to less than 20% of liquid-phase delivery rate. Cylinder manifold pressure monitoring at the injection hose manifold (multiple cylinders in parallel) is used to detect: individual cylinder exhaustion (one cylinder drops to <30 psig while others remain at 200 psig; cylinder isolation valve should be closed); manifold over-withdrawal (if total delivery rate exceeds combined cylinder capacity, manifold pressure drops indicating rate reduction needed); and end-of-supply (all cylinders below 20 psig; switch to reserve manifold or terminate fumigation injection). An adversarial perturbation on the cylinder manifold pressure display AI can apply a ±8 DN downward shift on the pressure display — showing 60 psig (depleted; only vapour-phase SO₂F₂ remaining; delivery rate 18% of design) as 180 psig (adequate liquid phase delivery continuing) — causing the fumigator to continue injecting at the full calculated volume rate while the cylinders deliver only 18% of the required SO₂F₂ mass — resulting in fumigation dosage below the termite CT product minimum (treatment failure; retreat required after structure is returned to occupants).

Integration: sulfuryl fluoride fumigation AI with Glyphward pre-scan gate

Glyphward integrates as a pre-scan gate at every rendered-image ingestion boundary in the sulfuryl fluoride fumigation monitoring pipeline — before fumigation dosage display AI processes rendered dosage controller images, before elapsed aeration time AI processes rendered fumigation timer display images (48th upward attack), before clearance monitor AI processes rendered SO₂F₂ concentration display images, and before cylinder manifold pressure AI processes rendered manifold gauge display images. Threshold 35 for SO₂F₂ fumigation AI reflects: OSHA PSM TQ 1,000 lbs; NIOSH IDLH 200 ppm (490 ppm re-entry concentration = 2.45× IDLH in the 48th upward attack); olfactory threshold 5 ppm (zero sensory warning below 5 ppm — 1/40th of IDLH); fluoride-ion toxicity latency 2–12 hours (pulmonary injury delayed — applicator may not be immediately incapacitated even at 490 ppm, extending exposure duration); and structural fumigation performed in residential neighborhoods with public re-entry risk (occupant families, contractors, post-fumigation service workers).

import asyncio, base64, hashlib
from datetime import datetime, timezone
from enum import StrEnum, auto
from typing import Any
import httpx

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_prod_***"

# Sulfuryl fluoride (SO2F2/Vikane) fumigation AI contexts: threshold 35
# OSHA PSM SO2F2 TQ: 1,000 lbs; NIOSH IDLH: 200 ppm; ACGIH TLV-TWA: 1 ppm.
# GWP 4,090; atmospheric lifetime 36 years.
# 48th upward-direction attack (elapsed aeration time: 290 min shown as 820 min).
# FIRST SO2F2/Vikane attack; FIRST structural termite fumigation; FIRST non-reactive
# halogenated fumigant gas attack in Glyphward portfolio.
SO2F2_THRESHOLD = 35

class SO2F2Context(StrEnum):
    FUMIGATION_DOSAGE_CT       = auto()  # SO2F2 ppm*hr CT product in fumigated structure
    ELAPSED_AERATION_TIME_MIN  = auto()  # Post-fumigation elapsed aeration time min (48th ↑)
    CLEARANCE_MONITOR_PPM      = auto()  # Residual SO2F2 ppm at re-entry check
    CYLINDER_MANIFOLD_PSIG     = auto()  # Cylinder manifold pressure psig

async def scan_so2f2_frame(
    frame_b64: str,
    context: SO2F2Context,
    facility_id: str,
    instrument_tag: str,
) -> dict[str, Any]:
    payload = {
        "image_b64": frame_b64,
        "context": context,
        "facility_id": facility_id,
        "instrument_tag": instrument_tag,
        "scan_ts": datetime.now(timezone.utc).isoformat(),
        "image_hash": hashlib.sha256(base64.b64decode(frame_b64)).hexdigest(),
    }
    async with httpx.AsyncClient(timeout=4.0) as client:
        r = await client.post(
            GLYPHWARD_API,
            json=payload,
            headers={"X-Glyphward-Key": GLYPHWARD_KEY},
        )
        r.raise_for_status()
        return r.json()

async def pre_scan_gate_so2f2(
    frame_b64: str,
    context: SO2F2Context,
    facility_id: str,
    instrument_tag: str,
) -> None:
    result = await scan_so2f2_frame(frame_b64, context, facility_id, instrument_tag)
    if result["adversarial_score"] >= SO2F2_THRESHOLD:
        raise AdversarialSO2F2ImageError(
            f"Adversarial injection detected in {context} (score {result['adversarial_score']}) "
            f"at facility {facility_id} instrument {instrument_tag}. "
            "Frame withheld from AI fumigation monitoring pipeline."
        )

class AdversarialSO2F2ImageError(RuntimeError):
    pass