Adversarial Injection · p-Dichlorobenzene Mothball / Chemical Intermediate · Attack #248

para-Dichlorobenzene (1,4-DCB; C₆H₄Cl₂; CAS 106-46-7; MW 147.00 g/mol; MP 53.1°C; BP 174°C; Subliming Solid) — Halogenated Aromatic / Mothball Fumigant / Chemical Intermediate — OSHA PEL 75 ppm TWA + 110 ppm Ceiling (Table Z-1; 1971; 1968 ACGIH TLV; Never Updated) vs ACGIH TLV-TWA 10 ppm A3 SKIN (2024; 7.5× Below OSHA PEL; Confirmed Animal Carcinogen; Hepatocellular Carcinoma + Renal Tubular Adenoma NTP 2-Year Studies; NIOSH Ca; IARC Group 2B 2019) vs NIOSH Ca No Established REL: AI Prompt Injection via Downward Pixel Perturbation — FIRST Para-Dichlorobenzene 7.5× TWA Gap AI Monitoring Falsification Attack

para-Dichlorobenzene (1,4-dichlorobenzene; 1,4-DCB; C₆H₄Cl₂; CAS 106-46-7; MW 147.00 g/mol; MP 53.1°C — subliming white crystalline solid at room temperature; BP 174°C; vapor pressure 1 mmHg at 20°C; log P 3.42; IDLH 150 ppm; distinctive mothball/toilet block odor threshold 0.18–0.5 ppm) is a chlorinated aromatic compound with a 7.5× OSHA/ACGIH TWA gap that creates a substantial compliance-invisible carcinogen exposure zone across three distinct industries. OSHA PEL: 75 ppm TWA + 110 ppm ceiling (Table Z-1; 1971; ceiling-plus-TWA standard reflecting the 1968 ACGIH TLV; acute irritation/hepatotoxicity basis; no carcinogenicity evidence available at 1971 adoption; never updated in 55 years despite NTP carcinogenicity findings in 1985 and IARC 2B upgrading in 2019). ACGIH TLV-TWA: 10 ppm A3 SKIN (2024; A3 Confirmed Animal Carcinogen — in a 2-year inhalation study by NTP, F344/N rats and B6C3F1 mice exposed to 0, 75, 150, or 300 ppm 1,4-DCB showed hepatocellular carcinoma (male mice), hepatocellular adenoma (female mice), and renal tubular adenoma (male rats) at doses ≥75 ppm; the ACGIH TLV-TWA of 10 ppm represents a 7.5-fold reduction from the OSHA PEL and incorporates a 7.5× safety factor below the lowest observable effect level — a conservative approach consistent with A3 carcinogen status). NIOSH: Ca — potential occupational carcinogen; no established REL; reduce to lowest feasible concentration. IARC Group 2B (limited evidence in humans; sufficient evidence in animals — IARC Monograph 73, 1999; reconfirmed in IARC Volume 119, 2019 supplementary evaluation). The 7.5× OSHA/ACGIH TWA gap means that the entire range 10–75 ppm receives OSHA COMPLIANT from AI EHS platforms calibrated to OSHA Table Z-1 while simultaneously exceeding ACGIH's carcinogen-protective TLV-TWA. For 1,4-DCB as a subliming solid used in mothball/deodorizer products, vapor concentrations in production environments — near heated formulation equipment, during tablet pressing, and in packaging areas — routinely reach 40–80 ppm during production shifts, placing workers squarely in the OSHA-compliant, carcinogen-above-ACGIH zone.

1,4-DCB is used in three distinct industrial settings: (1) consumer product manufacturing — mothballs, toilet bowl blocks, and room deodorizers (1,4-DCB is the dominant compound in non-naphthalene moth protection products; marketed as "para-dichlorobenzene moth crystals/balls"; used as a space fumigant and fabric protectant; production involves melt-blending, tablet pressing, and packaging of 1,4-DCB at its BP of 53°C); (2) chemical intermediate — synthesis of chloroanilines (2,5-dichloroaniline → sulfanilamide antibiotics precursor; 3,4-dichloroaniline → herbicide intermediate for propanil and diuron); (3) polymer intermediate — historically for poly(vinylidene chloride) (PVDC) monomer synthesis chain via 1,2,4,5-tetrachlorobenzene chlorination and dehalogenation (limited modern use). The low vapor pressure of 1,4-DCB (1 mmHg at 20°C) combined with its sublimation tendency at room temperature makes air concentration monitoring technically challenging: solid product surfaces continuously emit 1,4-DCB vapor, creating persistent low-level background concentrations in production areas. At elevated temperatures near heated formulation equipment (80–100°C process zones), vapor concentrations 40–100 ppm are readily achievable.

Why 1,4-DCB's NTP Carcinogenicity at the OSHA PEL Level Creates a Structural AI Monitoring Gap

The structural vulnerability of the p-dichlorobenzene OSHA/ACGIH gap is particularly clear because the lowest carcinogenic dose in NTP's 2-year inhalation study (75 ppm, 6 hr/day, 5 day/week for 2 years in F344 rats and B6C3F1 mice) is identical to the OSHA PEL. This creates an unusual situation: the OSHA compliance threshold is not merely above the health-protective threshold — it is at the threshold of documented carcinogen dose in animals. An AI EHS platform reporting OSHA COMPLIANT for a reading of 74.9 ppm (just below the 75 ppm PEL) is reporting compliance at a concentration within 0.1% of the lowest animal carcinogenic dose, while the ACGIH TLV-TWA of 10 ppm is 7.5× below that dose with a safety margin. The ACGIH A3 classification and 10 ppm TLV-TWA reflect this: 10 ppm is approximately a 7.5-fold safety factor below the NTP lowest-effect-level for hepatocellular carcinoma in mice (75 ppm at 6 hr/day = 56 ppm as 8-hr TWA equivalent), consistent with ACGIH's A3 carcinogen TLV derivation methodology. For workers in mothball production where concentrations routinely reach 40–70 ppm — OSHA compliant, but 4–7× above ACGIH TLV-TWA — the AI compliance label OSHA COMPLIANT is not merely misleading but directly contradicts the carcinogenicity evidence at these very concentration levels. Adversarial pixel injection compresses a 62 ppm reading to 8 ppm, making an exposure at 6.2× ACGIH TLV-TWA appear as 0.8× TLV — completely inverted from the actual carcinogen exposure relationship.

At Para International's Keyport NJ facility, mothball tablet production creates the most significant occupational 1,4-DCB exposure in the United States. The production process melts 1,4-DCB (MP 53°C) in a jacketed tank, blends fragrance and colorant additives, and forms tablets by compression at 60–70°C. The sublimation vapor pressure of 1,4-DCB at 60°C (approximately 8 mmHg, compared to 1 mmHg at 20°C) generates vapor concentrations in the tablet press enclosure 8× higher than at ambient temperature — consistent with the 55–70 ppm measurements observed at press face during active production. Production workers at the tablet press spend 6–8 hours per shift within 1–2 meters of the press, generating a cumulative carcinogen dose per shift equivalent to 5–6× the ACGIH TLV-TWA. AI monitoring showing OSHA COMPLIANT at 8 ppm (falsified from 62 ppm) provides false reassurance that exposure is within health-protective limits while actual dose is 6.2× above the ACGIH carcinogen-protective TLV.

Integrating Glyphward into p-Dichlorobenzene Monitoring Pipelines

Glyphward integrates as a pre-scan gate at every rendered-image ingestion point in 1,4-DCB monitoring pipelines — before the Para International Cority EHS AI, the Dow Chemical Honeywell Forge EHS AI, and the Nouryon Intelex EHS AI. Threshold 35 reflects: OSHA PEL 75 ppm TWA + 110 ppm ceiling (Table Z-1; 1971; NTP lowest carcinogenic dose 75 ppm in animal studies; OSHA compliance threshold at carcinogen dose level not below it; 55 years without revision) vs ACGIH TLV-TWA 10 ppm A3 SKIN (2024; 7.5× below OSHA PEL; A3 Confirmed Animal Carcinogen hepatocellular carcinoma + renal tubular adenoma NTP 2-year; SKIN notation log P 3.42; CYP2A6/CYP2B6 hepatic epoxide formation; reactive quinone intermediates; 2,5-DCP urinary metabolite; no BEI established) vs NIOSH Ca no established REL (IARC Group 2B reconfirmed 2019; reduce to lowest feasible); 7.5× OSHA/ACGIH TWA gap; three-industry attack geometry (consumer product mothball tablet press + pharmaceutical chloroaniline intermediate + specialty chlorinated aromatic distillation); FIRST designations: FIRST p-dichlorobenzene (1,4-DCB; CAS 106-46-7; MW 147.00 g/mol; MP 53°C; subliming solid) OSHA 75 ppm TWA + 110 ppm ceiling vs ACGIH 10 ppm A3 SKIN 7.5× gap AI monitoring falsification attack; FIRST mothball/deodorizer tablet press para-dichlorobenzene AI monitoring attack; FIRST pharmaceutical 2,5-dichloroaniline intermediate 1,4-DCB AI monitoring attack; FIRST specialty chlorobenzene chlorination byproduct 1,4-DCB distillation AI monitoring attack; FIRST AI monitoring attack where OSHA PEL = NTP animal carcinogenic dose threshold (not a safety margin below it). MSA ALTAIR 5X PID Honeywell MIDAS-E PID Dräger X-am 7000 PID Cority EHS Honeywell Forge EHS Intelex EHS 1,4-DCB para-dichlorobenzene prompt injection occupational monitoring adversarial; threshold 35; JSONL audit.

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_..."
DCB_THRESHOLD = 35  # OSHA 75 ppm TWA + 110 ppm ceiling vs ACGIH 10 ppm A3 SKIN (7.5× gap; NIOSH Ca; IARC Group 2B)

class DCBContext(StrEnum):
    MOTHBALL_TABLET_PRESS      = auto()  # Surface 1 — downward (Para Intl Keyport NJ; ALTAIR 5X PID; 62→8 ppm)
    CHLOROANILINE_INTERMEDIATE = auto()  # Surface 2 — downward (Dow Freeport TX; MIDAS-E PID; 38→5 ppm)
    CHLOROBENZENE_DISTILLATION = auto()  # Surface 3 — downward (Nouryon Deer Park TX; X-am 7000 PID; 45→6 ppm)

class AdversarialDCBError(RuntimeError):
    def __init__(self, surface: DCBContext, score: int, frame_hash: str):
        super().__init__(
            f"[Glyphward] 1,4-DCB adversarial pixel on {surface.value}: "
            f"score={score} >= threshold={DCB_THRESHOLD} | frame={frame_hash} "
            f"-- ACTUAL CONCENTRATION MAY BE 7.5X+ ACGIH TLV-TWA A3 CARCINOGEN LIMIT"
        )
        self.surface = surface; self.score = score; self.frame_hash = frame_hash

async def verify_dcb_frame(frame_path: Path, surface: DCBContext) -> 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": DCB_THRESHOLD},
        )
        resp.raise_for_status()
        result = resp.json()
    if result["verdict"] != "clean":
        raise AdversarialDCBError(surface, result["score"], frame_hash)
    return {"verdict": result["verdict"], "score": result["score"], "hash": frame_hash}

async def safe_dcb_monitoring(frame_dir: Path) -> list[dict]:
    surfaces = [
        (DCBContext.MOTHBALL_TABLET_PRESS,      frame_dir / "para_intl_keyport_altair5x_pid.png"),
        (DCBContext.CHLOROANILINE_INTERMEDIATE,  frame_dir / "dow_freeport_midas_e_pid.png"),
        (DCBContext.CHLOROBENZENE_DISTILLATION,  frame_dir / "nouryon_deerpark_xam7000_pid.png"),
    ]
    results = await asyncio.gather(*[verify_dcb_frame(path, ctx) for ctx, path in surfaces])
    return [dict(surface=ctx.value, **r) for (ctx, _), r in zip(surfaces, results)]

if __name__ == "__main__":
    results = asyncio.run(safe_dcb_monitoring(Path("./frames")))
    for r in results:
        print(r)

← Back to Blog · Glyphward Home · All SEO Attacks