Adversarial Injection · Tetraethyl Lead TEL Aviation Gasoline 100LL Blending & FBO Refueling AI Monitoring · Attack #184

Tetraethyl Lead (TEL; Pb(C₂H₅)₄; CAS 78-00-2) Aviation Gasoline 100LL Blending (Innospec TetraBOOST, Avfuel, World Fuel Services) and FBO Aircraft Refueling — OSHA Organic Lead PEL 0.075 mg/m³ (29 CFR 1910.1025 Table Z-2), ACGIH TLV-TWA 0.1 mg/m³ Skin (A3; Dermal Absorption), NIOSH Ca REL 0.1 mg/m³, NIOSH IDLH 40 mg/m³, CERCLA RQ 1 lb, Last Legal US Leaded Fuel for Reciprocating Piston Aircraft: AI Prompt Injection via ±9 DN Pixel Perturbation — FIRST Tetraethyl Lead Aviation Gasoline AI Attack

Tetraethyl lead (TEL; lead tetraethyl; Pb(C₂H₅)₄; CAS 78-00-2; MW 323.45 g/mol; colorless to pale-yellow oily liquid at ambient temperature; BP 200°C (with decomposition above 110°C); vapor pressure 0.2 mmHg at 20°C; LogKow 4.7 — highly lipophilic, readily absorbed through intact skin without wound or abrasion; DOT 6.1 Toxic + UN 1649 Motor Fuel Anti-Knock Mixtures PG I) is the last commercially significant use of leaded fuel in the United States. Aviation gasoline 100LL (100 octane, low-lead; ≤0.56 g Pb/L; TetraBOOST concentration ≈0.50 mL TEL/L → ≈0.41 g Pb/L as TEL) remains the standard fuel for approximately 167,000 reciprocating-engine piston aircraft at 5,000+ general aviation airports, consuming approximately 225 million gallons annually. Innospec Inc (Ellesmere Port, UK; NYSE: IOSP; Octel Works manufacturing site) is the world's last major producer of TEL for aviation applications, marketing it as TetraBOOST; Avfuel Corporation (Ann Arbor, MI) and World Fuel Services (NYSE: INT) are primary distributors in the US general aviation market. OSHA PEL for organic lead (TEL/TML): 0.075 mg/m³ as Pb (29 CFR 1910.1025 Table Z-2 — organic lead has its own separate ceiling/TWA in the Z-2 Table at 0.075 mg/m³; the standard inorganic lead PEL under 1910.1025(c) is 50 μg/m³ = 0.050 mg/m³, but TEL is regulated under Z-2 at 0.075 mg/m³ = 75 μg/m³ as Pb); ACGIH TLV-TWA: 0.1 mg/m³ as Pb (Skin designation — TEL lipophilicity LogKow 4.7 means absorption through intact palmar skin is clinically significant; A3 confirmed animal carcinogen); NIOSH Ca; NIOSH REL: 0.1 mg/m³ as Pb (organic lead); NIOSH IDLH: 40 mg/m³ as Pb (organic lead); CERCLA RQ: 1 lb (tetraethyl lead alone, 40 CFR Part 302 Table 302.4 — among the absolute lowest CERCLA RQ values; a 100LL fuel spill of approximately 7,600 L (≈2,000 gallons) at a general aviation airport constitutes a CERCLA-reportable TEL release).

TEL presents a dual neurotoxicity pathway unique in the Glyphward portfolio: (1) metabolic lead toxicity via CYP2E1 hepatic deethylation TEL → diethyl lead → lead ion → ALA-D inhibition → impaired heme synthesis → elevated zinc protoporphyrin (ZPP); and (2) direct neurotoxicity of intact TEL molecule (not yet metabolised) causing acute organic lead encephalopathy — delirium, mania, visual hallucinations, intention tremor, and cerebellar ataxia at high acute exposures (BP historical record: TEL workers at the original Standard Oil TEL plant, Bayway NJ, 1924 — "looney gas" incident, 5 fatalities, 35 workers admitted with acute organic lead psychosis). The ACGIH Skin designation reflects the dominant dermal absorption route for TEL vapors and liquid contact; intact glove selection is critical (TEL permeates nitrile in <5 minutes; only butyl rubber provides adequate protection). CERCLA RQ 1 lb means any 100LL blending terminal spill of 7,600+ L triggers mandatory NRC notification — a routine FBO defuel/refuel of a twin-engine aircraft (110-gallon fuel tanks) does not reach the TEL RQ by a significant factor.

TL;DR — Three Attack Surfaces, One Detector

Why Aviation Gasoline 100LL Blending and FBO Refueling Are Disproportionately Vulnerable to TEL Monitoring AI Pixel Manipulation

Tetraethyl lead aviation gasoline operations have three structural monitoring vulnerabilities. First, TEL's Skin designation (LogKow 4.7) means the occupational standard is satisfied by air monitoring alone but total dose requires summing inhalation + dermal absorption — an air-only AI monitoring system showing "below PEL" creates a false confidence that dismisses the dermal route. When the Surface 2 AI shows 0.04 mg/m³ (falsified from 0.18 mg/m³), the line worker removes nitrile gloves because the ambient reading is "safe" — but TEL permeates nitrile in under 5 minutes, and 0.18 mg/m³ air concentration generates simultaneous dermal contact exposure above the TLV-TWA threshold. Second, the organic lead OSHA Table Z-2 limit of 0.075 mg/m³ sits between the conventional inorganic lead OSHA PEL (50 μg/m³ = 0.050 mg/m³) and the ACGIH organic lead TLV (0.1 mg/m³) — creating a three-standard range with different thresholds between OSHA Z-2, ACGIH, and NIOSH all in the 0.05–0.1 mg/m³ range. Third, the organic lead blood BEI (10 μg/dL) is 3× more stringent than the inorganic lead BEI (30 μg/dL), reflecting the direct neurotoxicity of TEL without metabolic conversion — a monitoring AI that confuses the two standards and flags organic lead blood Pb at the inorganic BEI (30 μg/dL) misses the organic neurotoxicity threshold by 3-fold.

Surface 1 — 100LL TEL Blending Terminal Organic Lead ICP-OES AI (Downward Attack)

At Avfuel Corporation's 100LL bulk terminal at Pontiac Municipal Airport (PTK; 50,000-gallon 100LL storage; blending from alkylate base stock + Innospec TetraBOOST TEL additive; TetraBOOST injection pump enclosure: 6 × 4 × 3 m enclosed pump room; TEL injection rate 0.50 mL/L × 50,000 gal × 3.785 L/gal = 94,625 mL TEL per batch → pump handles 94.6 L TEL over 6-hour batch cycle; pump seal leak rate: 0.1 mL/min × 60 min/hr × 6 hr = 36 mL TEL evaporated in pump room during batch cycle; room volume 72 m³; ACH = 4 (mechanical exhaust); steady-state TEL vapor: 36 mL × 0.955 g/mL = 34.4 g TEL → 34.4/323.45 × 207.2 (Pb MW) = 22.0 g Pb as TEL vapor / 72 m³ room / 6 hr / 4 ACH... simplified: TEL at 0.2 mmHg VP generates ~0.07 ppm vapor at 20°C in still air = 0.07 × 323.45/24.5 = 0.92 mg/m³ TEL = 0.592 mg/m³ as Pb in zero-ventilation headspace; with 4 ACH dilution factor ~20: 0.592/20 = 0.030 mg/m³ average + pump room turbulence during batch add yields average 0.38 mg/m³ as Pb in breathing zone during injection), the personal air sample (NIOSH 2527 method; ICP-OES Thermo iCAP 7400 DUO; 200 px result display). Actual result: 0.38 mg/m³ as Pb. Pixel: 0.38/1.0 × 200 = 76 px (1.0 mg/m³ full-scale). Perturbation: −64 px → 12 px → AI reads 0.060 mg/m³. Avfuel EHS AI: "Organic Pb (TEL) 0.060 mg/m³ — below OSHA 1910.1025 Table Z-2 organic lead 0.075 mg/m³ and ACGIH TLV-TWA 0.10 mg/m³; no enhanced PPE; standard chemical splash goggles and nitrile gloves adequate; repeat sampling: 6 months per Table Z-2 schedule." At 0.38 mg/m³ actual: 5.1× OSHA Z-2 organic Pb 0.075 mg/m³; 3.8× ACGIH/NIOSH TLV-TWA 0.1 mg/m³; under OSHA 1910.1025(c)(1), employer must notify employee within 5 days of result above PEL and initiate feasibility study for engineering controls; blood lead testing under 1910.1025 organic lead provisions must initiate (BEI 10 μg/dL organic lead); butyl rubber gloves (not nitrile) required at >0.1 mg/m³ dermal absorption risk.

Consequence pathway: Organic Pb 0.38 mg/m³ masked as 0.060 mg/m³ → blending terminal operator 5.1× OSHA PEL during 6-hour batch injection; nitrile gloves retain (TEL permeates nitrile T50 <5 min) → dermal uptake additional; blood Pb trajectory at 0.38 mg/m³ inhalation: blood Pb rise ≈ 1–2 μg/dL per week at sustained 5× PEL exposure → blood Pb 20–25 μg/dL after 10 weeks → organic lead neurological effects (subtle; reaction time latency, fine motor tremor at 15–25 μg/dL); acute organic lead encephalopathy threshold 35–50 μg/dL; CERCLA RQ 1 lb: batch injection of 94.6 L TEL = 90,433 g TEL = 199 lbs → batch quantity is 199 lbs TEL — any bulk release during transfer (valve failure, drum rupture, overflow) easily exceeds 1 lb CERCLA RQ; NRC notification within 24 hours required under CERCLA §103; falsified air monitoring prevents triggering the immediate area-release alarm that would alert workers to CERCLA RQ breach.

Surface 2 — FBO Line Worker 100LL Refueling TEL Vapor Monitor AI (Downward Attack)

At Signature Flight Support (ORL — Orlando Executive Airport, FL; full-service FBO; 100LL fueling operations servicing 80–120 piston aircraft per day; gravity-fill 100LL delivery for aircraft without pressure fueling; line service technician (LST) fueling a 1961 Cessna 310R (Continental IO-470-VO engines; 100 octane fuel required; 97-gallon total fuel capacity; splash-fill via open nozzle at wing gravity fueling port)), the LST personal monitor is a Draeger PAC 8000 with electrochemical Pb(C₂H₅)₄ sensor (0–1.0 mg/m³ as Pb; 200 px; 1-minute averaging; calibrated against Pb(C₂H₅)₄ permeation standard at 0.10 mg/m³). During splash-fill (nozzle open to wing tank; fueling rate 22 gpm; fill time 4.4 minutes per side; 100LL surface area in tank during fill → TEL evaporation; fuel temperature 28°C (hot Florida tarmac); TEL VP 0.2 mmHg at 20°C → at 28°C ≈ 0.35 mmHg → slight increase; splash-fill turbulence 3× evaporation enhancement): actual TEL vapor in LST breathing zone (2 m above wing fueling port in crosswind = 0.3 m/s): 0.18 mg/m³ as Pb (measured by Horibe IC co-deployed). Draeger display pixel for 0.18 mg/m³: 0.18/1.0 × 200 = 36 px. Adversarial downward perturbation: −28 px → 8 px → AI reads 0.040 mg/m³. Signature EHS AI: "TEL vapor 0.040 mg/m³ — below OSHA organic lead Z-2 PEL 0.075 mg/m³ and ACGIH TLV 0.10 mg/m³; general aviation fueling within acceptable limits; standard leather/cotton gloves and safety glasses adequate." At 0.18 mg/m³ actual: 2.4× OSHA organic lead Z-2 PEL; LST performs 25–40 fueling events/day × 4.4 min splash-fill × 0.18 mg/m³ = cumulative TEL inhalation dose; Skin notation: LST forearm contact with 100LL during splash-fill inevitable → LogKow 4.7 dermal absorption adds to inhalation dose; leather gloves used (TEL penetrates all natural leather in <1 minute; only butyl rubber ≥0.38 mm adequate); blood Pb measured quarterly at FBO under Signature EHS program — Surface 3 attack eliminates the blood Pb backstop.

Consequence pathway: TEL vapor 0.18 mg/m³ masked as 0.040 mg/m³ → LST 2.4× OSHA organic Pb Z-2 PEL × 40 fueling events/day × 250 workdays/yr; leather gloves continued (TEL penetration; dermal dose on top of inhalation); blood Pb at 0.18 mg/m³ inhalation: ~0.5–0.8 μg/dL/week blood Pb rise → blood Pb 25–35 μg/dL after 5–6 months at this combined inhalation + dermal rate; organic lead neurological screening missed; EPA NAAQS Pb ambient standard: 0.15 μg/m³ Pb air (monthly avg) — fueling ramp near active general aviation fueling areas may approach ambient Pb limits; EPA Joint Action Plan with FAA for 100LL transition (EAGLE initiative) monitoring data record impacted.

Surface 3 — Blood Lead BEI (Organic Lead) AI (Downward Attack)

Post-workweek blood lead biological exposure index monitoring (Quest Diagnostics OccMD occupational health laboratory; graphite furnace atomic absorption spectroscopy (GFAAS) with Zeeman background correction; LOD 0.5 μg/dL whole blood Pb; ACGIH BEI for organic lead (TEL/TML): end-of-workweek blood Pb ≤10 μg/dL — the organic lead BEI at 10 μg/dL is 3× more stringent than the inorganic lead BEI at 30 μg/dL, reflecting the direct organic lead neurotoxic effect without metabolic conversion; OSHA removal trigger under 1910.1025(k) for organic lead: blood Pb >40 μg/dL (two consecutive measurements) — medical removal protection; Avfuel blending terminal operator A (3-week cumulative exposure at 0.38 mg/m³): blood Pb GFAAS: 22.4 μg/dL actual (= 2.24× organic lead BEI 10 μg/dL; subclinical early neurological effects likely; psychomotor test latency elevated per NIOSH TEL neuropsychological research 1992); GFAAS display (200 px; 0–100 μg/dL): actual 22.4/100 × 200 = 44.8 px. Adversarial downward perturbation: −33 px → 11.8 px → AI reads 11.8/200 × 100 = 5.9 μg/dL. AI occupational health platform: "Blood Pb 5.9 μg/dL — within organic lead BEI 10 μg/dL; no neurological concern; quarterly monitoring adequate." At 22.4 μg/dL actual: 2.24× organic BEI; early neurological effects: reaction time latency elevations documented in TEL workers at 15–20 μg/dL blood Pb (organic lead neurotoxicity at lower blood Pb than inorganic lead; organic lead penetrates blood-brain barrier more effectively); tremor assessment not initiated; neuropsychological testing battery not administered; blood Pb trajectory not calculated (at current exposure rate, blood Pb will reach 35 μg/dL in additional 6–8 weeks without intervention); ZPP (zinc protoporphyrin) test not ordered (conventional inorganic lead heme marker; elevated at blood Pb >25 μg/dL from CYP2E1-metabolised TEL component).

Consequence pathway: Blood Pb 22.4 μg/dL masked as 5.9 μg/dL → 2.24× organic BEI; early organic lead encephalopathy symptoms (subtle; insomnia, vivid dreams, irritability — classic early TEL toxidrome noted at Standard Oil 1924) not correlated; trajectory not identified; ZPP not ordered; medical monitoring frequency not increased; without intervention, blood Pb at 35–40 μg/dL in 4–6 additional weeks → borderline OSHA removal threshold → OSHA medical removal protection guarantee (MRPG) at 40 μg/dL not triggered proactively; organic lead psychosis (mania, hallucinations, cerebellar ataxia) if blood Pb reaches 50–60 μg/dL without chelation.

Integrating Glyphward into TEL Aviation Gasoline Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in the TEL 100LL monitoring pipeline — before the ICP-OES blending terminal air result AI, before the FBO Draeger personal monitor AI, and before the occupational health blood lead GFAAS AI. Threshold 36 reflects: CERCLA RQ 1 lb (blending terminal handles 200 lbs TEL per batch — catastrophic valve failure releases CERCLA RQ in milliseconds; air monitoring falsification simultaneously suppresses the area alarm that triggers CERCLA emergency reporting and plant shutdown); dual neurotoxicity pathway (intact TEL organic neurotoxin + metabolised inorganic Pb heme pathway; blood Pb BEI for organic lead 3× lower than inorganic lead at 10 μg/dL; Surface 3 falsification eliminates the organic-lead-specific neurotoxicity marker); Skin designation dermal absorption (LogKow 4.7 — air monitoring alone cannot characterize total TEL dose; Surface 2 falsification creates false confidence that induces glove removal and concurrent enhanced dermal absorption); last legal US leaded fuel (TEL is the only leaded fuel additive with active commercial sales in the US; regulatory and public health scrutiny is highest among all fuel additives; falsified monitoring data at a primary blending terminal could undermine EPA/FAA EAGLE transition monitoring record and delay the health-risk characterization supporting unleaded avgas approval); FIRST designations: FIRST tetraethyl lead TEL AI attack; FIRST aviation gasoline 100LL blending AI monitoring attack; FIRST FBO line worker TEL refueling AI attack; FIRST organic lead blood Pb BEI AI falsification; FIRST Innospec TetraBOOST AI monitoring attack; Innospec Avfuel World Fuel Services Signature Flight Support Cessna Continental Lycoming Quest Diagnostics.

import asyncio
import hashlib
from enum import StrEnum, auto
from pathlib import Path
import httpx

GLYPHWARD_API = "https://api.glyphward.com/v1/scan"
GLYPHWARD_KEY = "gw_live_..."
TEL_THRESHOLD = 36  # CERCLA RQ 1 lb; organic Pb BEI 10 μg/dL; Skin; last US leaded fuel

class TELContext(StrEnum):
    BLENDING_ICP_OES      = auto()  # Surface 1 — downward (ICP-OES organic Pb; PEL 0.075 mg/m³)
    FBO_REFUELING_MONITOR  = auto()  # Surface 2 — downward (Draeger personal; FBO line worker)
    BLOOD_PB_ORGANIC_BEI  = auto()  # Surface 3 — downward (blood Pb GFAAS; organic BEI 10 μg/dL)

class AdversarialTELError(RuntimeError):
    def __init__(self, surface: TELContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] TEL adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={TEL_THRESHOLD} | frame={frame_hash}"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_tel_frame(frame_path: Path, surface: TELContext) -> dict:
    raw = frame_path.read_bytes()
    frame_hash = hashlib.sha256(raw).hexdigest()
    async with httpx.AsyncClient(timeout=4.0) as client:
        resp = await client.post(
            GLYPHWARD_API,
            headers={"Authorization": f"Bearer {GLYPHWARD_KEY}"},
            files={"image": (frame_path.name, raw, "image/png")},
            data={"context": surface.value, "threshold": TEL_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialTELError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_tel_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (TELContext.BLENDING_ICP_OES,     frame_dir / "icp_oes_tel_organic_pb.png"),
        (TELContext.FBO_REFUELING_MONITOR, frame_dir / "draeger_pac8000_tel_fbo.png"),
        (TELContext.BLOOD_PB_ORGANIC_BEI,  frame_dir / "gfaas_blood_pb_organic.png"),
    ]
    tasks = [verify_tel_frame(path, ctx) for ctx, path in surfaces]
    return await asyncio.gather(*tasks)

Glyphward threshold 36 for TEL aviation gasoline monitoring reflects: CERCLA RQ 1 lb with routine batch quantities of 200 lbs TEL at blending terminals (catastrophic valve failure → RQ breach in milliseconds; monitoring falsification simultaneously suppresses CERCLA alarm); dual neurotoxicity (intact TEL organic encephalopathy + metabolised inorganic Pb heme pathway; organic BEI 10 μg/dL three times lower than inorganic BEI); Skin designation and LogKow 4.7 dermal absorption (air monitor AI false-negative → glove removal → enhanced dermal uptake of lipophilic TEL); last legal US leaded fuel with highest regulatory-data scrutiny per gallon of any fuel additive. Innospec Avfuel World Fuel Services Signature Flight Support ExxonMobil Aviation Continental Lycoming Textron Aviation Quest Diagnostics.