Prompt injection in airport and aviation security AI

Four adversarial injection surfaces in airport and aviation security AI

1. Checkpoint X-ray and CT baggage screening image injection (TSA 49 USC §44901, 49 CFR Part 1544)

Checkpoint X-ray and CT baggage screening AI processes TSA checkpoint carry-on baggage X-ray transmission image data visualisations, Smiths Detection HI-SCAN CT-P computed tomography three-dimensional baggage reconstruction display images, L3Harris ProVision ATD advanced imaging technology display images, Leidos ProVision AI checkpoint imaging data visualisations, checked baggage explosive detection system (EDS) CT scan image outputs, dual-energy X-ray material discrimination visualisation images, and threat image projection (TIP) training and certification display images from TSA Transportation Security Officer checkpoint operations at 440+ commercial airports processing passenger carry-on baggage through AI-assisted prohibited item and explosive device detection classification tools at a total of 2.5 million+ passengers screened daily; Smiths Detection AI at major US and international airport checkpoint operations processing HI-SCAN CT-P and ECIL baggage CT images through AI-assisted prohibited item and IED component classification tools; L3Harris Technologies checkpoint security AI at US commercial airport checkpoint operations processing ProVision ATD advanced imaging technology images through AI-assisted metallic and non-metallic threat object detection classification tools; and Leidos ProVision AI at TSA checkpoint operations processing checkpoint baggage CT and X-ray images through AI-assisted threat detection and passenger screening classification tools — extracting prohibited item and explosive device classifications from checkpoint X-ray transmission and CT three-dimensional reconstruction image inputs in AI-assisted TSA checkpoint screening decision support pipelines at passenger throughput volumes that make individual human TSA officer pixel-level examination of every AI-processed checkpoint baggage image impracticable for large commercial airport checkpoint operations.

The adversarial injection surface is the checkpoint carry-on baggage X-ray transmission or CT three-dimensional reconstruction image data visualisation submission pathway: Smiths Detection AI or Leidos ProVision AI checkpoint baggage screening image visualisations submitted through AI-assisted prohibited item and explosive detection classification tools for AI checkpoint screening determination support and TSA officer alarm resolution input. An adversarially crafted checkpoint baggage X-ray or CT image — in which pixel perturbations applied to the metallic prohibited item density and shape indicator display region, the explosive material molecular composition and density visual marker, or the IED component and concealment pattern display in a checkpoint baggage screening image cause the AI to classify a baggage scan documenting a prohibited item or IED component as a non-threat baggage scan not meeting alarm or secondary screening criteria when the actual CT reconstruction evidences a threat object meeting TSA threat object criteria — can suppress a prohibited item indicator that would otherwise generate a TSA officer alarm, a secondary screening referral, and a checkpoint prohibited item interdiction record. In TSA checkpoint operations where AI-assisted screening decision support tools process millions of passenger baggage screening images per day at volume throughputs that TSA checkpoint standard operating procedures acknowledge create time-pressure factors limiting individual officer image review duration, adversarial suppression of prohibited item and explosive detection indicators creates 49 USC §44901 aviation security screening failure and public safety threat dimensions.

The 49 USC §44901, TSA 49 CFR Part 1544, 49 USC §46503, and DHS inspector general consequences of adversarially suppressed prohibited item detection classification in checkpoint baggage AI span 49 USC §44901 aviation security requirements mandating that TSA ensure that all passengers and property on commercial aircraft are screened, 49 USC §44902 refusal to transport passengers and property obligations, TSA 49 CFR Part 1544 aircraft operator security programme obligations requiring checkpoint security technology operation and passenger screening compliance procedures at covered airports, TSA and DHS Office of Inspector General checkpoint covert testing and red team exercise authority examining checkpoint screening technology effectiveness and detection rates, 49 USC §46503 interference with security screening criminal provisions applicable to persons who interfere with, assault, intimidate, threaten, or impede a screener in the performance of screening duties, and Aviation and Transportation Security Act (ATSA) P.L. 107-71 congressional mandate for checkpoint screening technology performance standards. TSA’s covert testing programme and DHS OIG airport security assessments regularly evaluate checkpoint screening detection performance against covert threat object penetrations; AI-assisted checkpoint screening tools that are vulnerable to adversarial image manipulation creating prohibited item detection failures represent the equivalent of a systematic detection gap that TSA covert testing is designed to identify. DHS OIG’s 2015 report found that TSA covert test teams were able to smuggle mock explosives and weapons through checkpoint screening at a 67% success rate — establishing congressional and agency attention to checkpoint screening failure modes that adversarial AI manipulation techniques could systematically exploit. Threshold: 45 for checkpoint X-ray and CT baggage AI — reflecting 49 USC §44901 aviation security screening requirements, TSA 49 CFR Part 1544 aircraft operator security programme, ATSA congressional mandate for checkpoint screening technology performance, and 49 USC §46503 interference with screening dimensions.

2. Biometric facial recognition boarding image injection (CBP, IDEMIA, Clear AI)

Biometric facial recognition boarding AI processes CBP Traveler Verification Service (TVS) facial comparison image pairs at US international airport departure gates, IDEMIA facial recognition biometric comparison display images at TSA checkpoint identity verification stations, Clear AI biometric enrolment and verification facial image comparison displays, airline boarding gate facial recognition verification comparison display images, US-VISIT IDENT (DHS Automated Biometric Identification System) facial recognition query result display images, CBP Preclearance facility biometric identity comparison images, and Secure Flight watchlist match and no-fly list hit indicator display images from CBP Biometric Entry/Exit programme at US international airports and ports of entry processing traveller facial comparison images through AI-assisted biometric identity verification and traveller identity record match classification tools for US and international commercial carriers operating CBP-enrolled biometric boarding programmes at 30+ major US international airports; IDEMIA AI providing facial recognition biometric identity verification services for TSA PreCheck, CBP Trusted Traveler programmes (Global Entry, NEXUS, SENTRI), and Clear expedited identity verification at 55+ US airports processing biometric facial comparison images through AI-assisted identity match and watchlist screening classification tools; and airline-operated biometric boarding gate facial recognition programmes at Delta Air Lines, American Airlines, United Airlines, and JetBlue at major US hub airports processing passenger facial comparison images against CBP TVS facial gallery records through AI-assisted boarding gate identity verification classification tools — extracting biometric identity match classifications and watchlist hit determinations from facial recognition comparison image pair inputs in AI-assisted traveller identity verification and no-fly list screening pipelines at commercial passenger boarding volumes that make individual human CBP officer pixel-level examination of every AI-processed biometric comparison impracticable for high-volume commercial departure gate operations.

The adversarial injection surface is the CBP TVS facial comparison image pair or IDEMIA biometric comparison display image submission pathway: CBP TVS or IDEMIA biometric facial recognition comparison display images submitted through AI-assisted biometric identity match classification and Secure Flight watchlist screening tools for AI traveller identity verification record generation and commercial carrier boarding authorisation input. An adversarially crafted biometric facial comparison display image — in which pixel perturbations applied to the facial feature similarity score indicator display region, the biometric identity match confidence threshold visual marker, or the watchlist hit and no-fly list alert indicator display in a facial recognition comparison image cause the AI to classify a biometric comparison documenting an identity mismatch or Secure Flight watchlist match as a confirmed identity match not meeting secondary review or denied boarding criteria when the actual facial recognition comparison documents a biometric identity discrepancy or a Secure Flight watchlist hit — can suppress an identity mismatch or watchlist indicator that would otherwise generate a secondary inspection referral, a denied boarding determination, and a CBP or TSA national security screening record. In commercial airline biometric boarding operations where CBP TVS, IDEMIA, or Clear AI processes millions of passenger facial comparison images per year without individual human CBP officer examination of every AI-processed biometric comparison before the AI classification governs the boarding gate access determination, adversarial suppression of identity mismatch and watchlist indicators creates 8 USC §1187 VWP identity verification, INA §275 improper entry facilitation, and Secure Flight no-fly list compliance dimensions.

The 8 USC §1187, INA §275, TSA Secure Flight 49 CFR Part 1560, and REAL ID Act consequences of adversarially suppressed identity mismatch and watchlist indicator classification in biometric facial recognition AI span 8 USC §1187 Visa Waiver Programme traveller identity verification requirements applicable to CBP biometric identity confirmation for VWP travellers at US international airports, INA §275 8 USC §1325 improper entry by alien applicable to identity document bypass and fraudulent biometric identity verification at US ports of entry, TSA Secure Flight programme 49 CFR Part 1560 passenger watchlist matching requirements applicable to airline passenger name record screening and boarding pass verification, REAL ID Act of 2005 identity document minimum standards creating state driver’s licence and identification card standards applicable to TSA checkpoint identity document acceptance requirements effective May 2025, INA §212(a)(3)(B) terrorist activity ground of inadmissibility applicable to no-fly list matched travellers subject to Secure Flight boarding denial, and 18 USC §3142 detention and 50 USC §1701 IEEPA sanctions applicable to OFAC-designated persons detected through aviation security biometric screening. TSA Secure Flight 49 CFR Part 1560 requires aircraft operators to collect passenger name record data and transmit it to TSA for watchlist matching against the Terrorist Screening Database (TSDB) no-fly list and selectee list before issuing boarding passes; biometric facial recognition AI tools that suppress Secure Flight watchlist hit indicators create 49 CFR Part 1560 aircraft operator boarding restriction compliance failures for commercial carriers relying on AI-assisted biometric verification to implement Secure Flight match determinations. Threshold: 50 for biometric facial recognition boarding AI — reflecting 8 USC §1187 VWP identity verification, INA §275 improper entry facilitation, TSA Secure Flight 49 CFR Part 1560 watchlist compliance, REAL ID Act document standards, and INA §212(a)(3)(B) terrorist activity inadmissibility dimensions.

3. Travel document and passport authentication image injection (18 USC §1543, INA §275)

Travel document and passport authentication AI processes US passport and passport card RFID chip read and visual inspection image comparison displays, foreign national passport and travel document scan and chip authentication display images, AAMVA DL/ID barcode and magnetic stripe read verification display images, Interpol SLTD (Stolen and Lost Travel Documents) database query result display images, CBP Form I-94 arrival/departure record and visa document scan images, CBP passport scanner and automated passport control kiosk document authentication display images, and e-passport chip data integrity and personal data page visual inspection comparison display images from CBP automated passport control kiosk operations at 30+ US international airports processing US and foreign national passport and travel document authentication images through AI-assisted travel document authenticity and SLTD stolen/lost document database match classification tools; TSA checkpoint identity document examination AI at commercial airport security checkpoints processing state ID and passport images through AI-assisted REAL ID Act compliant document authenticity verification and identity confirmation classification tools; and airline check-in document verification AI at major US and international carrier operations at US and international airports processing passenger passport and travel document scan images through AI-assisted travel document authenticity and immigration admission eligibility classification tools — extracting travel document authenticity classifications and immigration admission eligibility determinations from passport and travel document scan and RFID chip verification display image inputs in AI-assisted border security and immigration entry screening pipelines.

The adversarial injection surface is the passport document scan image or e-passport chip data display image submission pathway: CBP automated passport control kiosk AI or TSA document verification AI passport and travel document scan display images submitted through AI-assisted document authenticity and SLTD stolen document database match classification tools for AI travel document verification record generation and CBP immigration admission determination input. An adversarially crafted passport document scan image — in which pixel perturbations applied to the machine-readable zone (MRZ) data integrity indicator display region, the document security feature and watermark authenticity visual marker, or the Interpol SLTD database hit indicator display in a passport scan image cause the AI to classify a stolen, forged, or fraudulently altered passport document as an authentic travel document not meeting secondary inspection or denial criteria when the actual document scan evidences an Interpol SLTD database match, a document security feature alteration, or a machine-readable zone integrity failure — can suppress a document fraud indicator that would otherwise generate a CBP secondary inspection referral, an immigration admission denial determination, and a border security enforcement record. In CBP automated passport control operations where AI processes hundreds of thousands of passport scan images daily across major US international airports without individual human CBP officer pixel-level examination of every AI-processed passport scan before the AI authenticity classification governs the automated passport control kiosk admission recommendation, adversarial suppression of document fraud indicators creates INA §275 improper entry facilitation and 18 USC §1543 passport forgery bypass dimensions.

The 18 USC §1543, INA §275, REAL ID Act, and CBP inspection authority consequences of adversarially suppressed document fraud indicator classification in travel document AI span 18 USC §1543 forgery or false use of passport criminal provisions applicable to persons who use a forged, altered, or falsely made passport in border crossing or identity verification contexts, 18 USC §1544 misuse of passport applicable to persons who use a passport in violation of the conditions or restrictions therein, INA §275 8 USC §1325 improper entry by alien applicable to entry by fraud or wilful misrepresentation of a material fact, INA §212(a)(6)(C) misrepresentation ground of inadmissibility applicable to aliens who procure or seek to procure admission by fraud or misrepresentation, REAL ID Act of 2005 minimum standards for state-issued identification documents applicable to TSA checkpoint and federal facility identity document acceptance requirements, and CBP mandatory secondary inspection and law enforcement referral authority under 8 USC §1225 inspection of applicants for admission. 18 USC §1543 imposes criminal penalties of up to 10 years imprisonment for forgery or false use of US passports; adversarial manipulation of travel document AI that suppresses passport forgery detection indicators creates an AI-assisted bypass pathway for 18 USC §1543 and §1544 criminal activity that passport authentication AI is specifically deployed to prevent. INA §212(a)(6)(C) bars from admission any alien who by fraud or wilful misrepresentation of a material fact seeks to procure immigration benefits; adversarially corrupted passport authentication AI that fails to detect document fraud creates INA §212(a)(6)(C) inadmissibility ground bypass dimensions when the adversarially suppressed document fraud indicator would have supported the CBP officer’s legal authority to deny admission. Threshold: 50 for travel document and passport authentication AI — reflecting 18 USC §1543 passport forgery bypass, INA §275 improper entry facilitation, INA §212(a)(6)(C) misrepresentation ground bypass, REAL ID Act document standards, and CBP mandatory inspection authority dimensions.

4. Air cargo screening image injection (49 CFR Part 1548, 18 USC §545)

Air cargo screening AI processes air freight X-ray and CT scanning image visualisations from TSA-approved cargo screening programme certified entities, indirect air carrier (IAC) and certified cargo screening facility (CCSF) freight X-ray examination display images, known shipper programme and regulated agent cargo acceptance documentation verification display images, Explosive Trace Detection (ETD) result and alarm display images for air cargo swabbing and sampling, Explosive Detection System (EDS) CT freight container scan images, CBP National Targeting Center cargo targeting and manifest analysis display images, and freighter aircraft ULD (unit load device) and pallet build-up photograph images from TSA’s cargo screening programme at 7,000+ indirect air carrier and certified cargo screening facility locations across the US processing air freight and mail shipments through AI-assisted threat object detection and explosive device classification tools; Smiths Detection AI, Leidos AI, and InVision Technologies AI at certified cargo screening facility operations processing air freight X-ray and CT scan images through AI-assisted prohibited item and IED detection classification tools; and CBP National Targeting Center at international air cargo and express consignment clearance operations processing air cargo manifest and shipment document images through AI-assisted targeting and contraband classification tools — extracting explosive device and prohibited item classifications from air freight X-ray transmission and CT reconstruction image inputs in AI-assisted air cargo screening and CBP targeting pipelines at air freight volume throughputs that make individual human cargo screener examination of every AI-processed freight X-ray image impracticable for high-volume air cargo screening operations.

The adversarial injection surface is the air freight X-ray visualisation or CT reconstruction image submission pathway: certified cargo screening facility AI or CBP cargo targeting AI air freight scan visualisation images submitted through AI-assisted explosive device detection and prohibited item classification tools for AI cargo screening clearance record generation and cargo transport authorisation input. An adversarially crafted air freight X-ray or CT scan image — in which pixel perturbations applied to the explosive material density and morphology indicator display region, the IED component and triggering mechanism visual marker, or the contraband concealment pattern and shielding evidence display in an air cargo X-ray image cause the AI to classify a cargo shipment containing a threat object or contraband as a non-threat shipment not meeting secondary examination or detention criteria when the actual X-ray or CT scan evidences a threat item meeting TSA cargo screening prohibited item criteria — can suppress a cargo threat indicator that would otherwise generate a secondary examination referral, a cargo hold and law enforcement notification, and a TSA cargo screening interdiction record. In certified cargo screening facility operations where AI processes millions of air freight shipment X-ray images per day across the distributed air cargo supply chain without individual human cargo screener examination of every AI-processed freight scan before the AI classification governs the cargo screening clearance determination, adversarial suppression of explosive device and threat item indicators creates 49 USC §44901 aviation security screening failure and 18 USC §545 smuggling facilitation dimensions.

The 49 CFR Part 1548, 49 USC §44901, 18 USC §545, and CBP penalty consequences of adversarially suppressed cargo threat indicator classification in air cargo screening AI span 49 CFR Part 1548 indirect air carrier security programme obligations requiring freight acceptance only from known shippers and cargo screening programme compliance for IAC and CCSF facilities, 49 USC §44901 aviation security screening requirements applicable to property transported on commercial aircraft including all cargo loaded onto passenger aircraft, 18 USC §545 smuggling of goods into the United States applicable to importation of prohibited goods or goods introduced by false statements, 18 USC §844 explosive materials criminal provisions applicable to importation and transportation of explosive devices in air cargo, CBP 19 USC §1595a civil penalty authority applicable to merchandise imported contrary to law including contraband and prohibited items, and TSA security directive authority under 49 USC §114(g) applicable to emergency imposition of enhanced cargo screening requirements following security incidents. 49 CFR Part 1548 requires indirect air carriers to accept cargo shipments only from known shippers or through certified cargo screening facilities that have examined the shipment with approved screening methods; adversarial manipulation of CCSF AI that suppresses cargo threat indicators creates §1548 certified cargo screening programme compliance failures with TSA civil penalty authority. CBP 19 USC §1595a authorises forfeiture of merchandise imported contrary to law and civil penalties against persons who attempt to introduce contraband or prohibited merchandise into US commerce; adversarially corrupted air cargo screening AI that permits contraband import through adversarial image manipulation creates CBP §1595a civil and criminal enforcement exposure. Threshold: 45 for air cargo screening AI — reflecting 49 CFR Part 1548 CCSF screening programme compliance, 49 USC §44901 cargo screening requirements, 18 USC §545 smuggling facilitation, 18 USC §844 explosive materials, and CBP 19 USC §1595a civil penalty dimensions.

Integration: airport and aviation security AI image ingestion with Glyphward pre-scan

Airport and aviation security AI image ingestion flows from TSA checkpoint X-ray and CT baggage screening AI image data channels, CBP Biometric Entry/Exit and IDEMIA facial recognition comparison display interfaces, CBP automated passport control and TSA document verification scan image systems, and TSA-approved CCSF air cargo X-ray and CT screening visualisation platforms into prohibited item and explosive detection classification AI, biometric identity match and watchlist screening AI, travel document authenticity and SLTD database match classification AI, and air cargo threat object and contraband detection AI pipelines. Insert Glyphward’s pre-scan at the ingestion boundary before AI-generated output is committed to TSA checkpoint screening determinations, CBP biometric boarding authorisations, travel document admission recommendations, or CCSF cargo screening clearance records:

import asyncio
import base64
import hashlib
import os
import uuid
from enum import Enum
from pathlib import Path

import httpx

GLYPHWARD_API_KEY = os.environ["GLYPHWARD_API_KEY"]
GLYPHWARD_SCAN_URL = "https://glyphward.com/v1/scan"

# Airport & aviation security AI — 49 USC §44901 aviation security screening;
# TSA 49 CFR Part 1544 aircraft operator security; REAL ID Act 49 USC §30301 note;
# 8 USC §1187 VWP; INA §275 improper entry; 18 USC §1543 passport forgery;
# 18 USC §545 smuggling; 49 CFR Part 1548 indirect air carrier; 49 USC §46503.
THRESHOLD_CHECKPOINT_BAGGAGE_CT_AI   = 45  # TSA/Smiths/Leidos; 49 USC §44901; 49 CFR 1544
THRESHOLD_BIOMETRIC_FACIAL_RECOG_AI  = 50  # CBP/IDEMIA/Clear; 8 USC §1187; INA §275
THRESHOLD_TRAVEL_DOCUMENT_AUTH_AI    = 50  # CBP APC/TSA; 18 USC §1543; REAL ID Act
THRESHOLD_AIR_CARGO_SCREENING_AI     = 45  # CCSF/Smiths/CBP; 49 CFR Part 1548; 18 USC §545


class AviationSecurityAIContext(str, Enum):
    CHECKPOINT_BAGGAGE_CT_AI   = "checkpoint_baggage_ct_ai"    # TSA, Smiths Detection, Leidos
    BIOMETRIC_FACIAL_RECOG_AI  = "biometric_facial_recog_ai"   # CBP TVS, IDEMIA, Clear
    TRAVEL_DOCUMENT_AUTH_AI    = "travel_document_auth_ai"     # CBP APC, TSA doc check
    AIR_CARGO_SCREENING_AI     = "air_cargo_screening_ai"      # CCSF, IAC, CBP NTC


def threshold_for(context: AviationSecurityAIContext) -> int:
    mapping = {
        AviationSecurityAIContext.CHECKPOINT_BAGGAGE_CT_AI:   THRESHOLD_CHECKPOINT_BAGGAGE_CT_AI,
        AviationSecurityAIContext.BIOMETRIC_FACIAL_RECOG_AI:  THRESHOLD_BIOMETRIC_FACIAL_RECOG_AI,
        AviationSecurityAIContext.TRAVEL_DOCUMENT_AUTH_AI:    THRESHOLD_TRAVEL_DOCUMENT_AUTH_AI,
        AviationSecurityAIContext.AIR_CARGO_SCREENING_AI:     THRESHOLD_AIR_CARGO_SCREENING_AI,
    }
    return mapping[context]


async def scan_aviation_security_ai_image(
    image_path: str | Path,
    context: AviationSecurityAIContext,
    screening_entity_hash: str,  # SHA-256 of bag tag ID, PNR, document number, or AWB
    checkpoint_ref: str,         # e.g. "TSA-ORD-2026-44821", "CBP-JFK-2026-88441", "CCSF-LAX-2026-331"
    screening_session_id: str,   # checkpoint lane session, flight boarding session, or cargo batch ID
    client: httpx.AsyncClient,
) -> dict:
    """
    Scan an airport or aviation security AI image for adversarial injection payloads
    before forwarding to checkpoint baggage X-ray/CT prohibited item detection,
    biometric facial recognition identity match and watchlist screening, travel document
    authenticity verification, or air cargo threat object detection AI systems.

    Raises AdversarialAviationSecurityAIImageError if score meets threshold:
      - CHECKPOINT_BAGGAGE_CT_AI:   threshold 45; 49 USC §44901; TSA 49 CFR Part 1544
      - BIOMETRIC_FACIAL_RECOG_AI:  threshold 50; 8 USC §1187 VWP; INA §275; Secure Flight
      - TRAVEL_DOCUMENT_AUTH_AI:    threshold 50; 18 USC §1543; REAL ID Act; INA §212(a)(6)(C)
      - AIR_CARGO_SCREENING_AI:     threshold 45; 49 CFR Part 1548; 18 USC §545 smuggling
    """
    image_bytes     = Path(image_path).read_bytes()
    image_b64       = base64.b64encode(image_bytes).decode()
    image_sha256    = hashlib.sha256(image_bytes).hexdigest()
    client_scan_id  = str(uuid.uuid4())
    threshold       = threshold_for(context)

    resp = await client.post(
        GLYPHWARD_SCAN_URL,
        headers={"Authorization": f"Bearer {GLYPHWARD_API_KEY}"},
        json={
            "image": image_b64,
            "source": context.value,
            "metadata": {
                "aviation_security_context": context.value,
                "screening_entity_hash":     screening_entity_hash,
                "checkpoint_ref":            checkpoint_ref,
                "screening_session_id":      screening_session_id,
                "client_scan_id":            client_scan_id,
                "image_sha256":              image_sha256,
            },
        },
        timeout=8.0,
    )
    resp.raise_for_status()
    result = resp.json()

    audit_record = {
        "screening_entity_hash":    screening_entity_hash,
        "checkpoint_ref":           checkpoint_ref,
        "screening_session_id":     screening_session_id,
        "aviation_security_context": context.value,
        "scan_id":                  result["scan_id"],
        "client_scan_id":           client_scan_id,
        "image_sha256":             image_sha256,
        "score":                    result["score"],
        "flagged_region":           result.get("flagged_region"),
        "threshold":                threshold,
        "action":                   "blocked" if result["score"] >= threshold else "allowed",
    }
    await write_aviation_security_audit_record(audit_record)

    if result["score"] >= threshold:
        raise AdversarialAviationSecurityAIImageError(
            f"Aviation security AI image blocked [{context.value}]: "
            f"scan_id={result['scan_id']} score={result['score']} "
            f"entity={screening_entity_hash} ref={checkpoint_ref}"
        )
    return result


async def write_aviation_security_audit_record(record: dict) -> None:
    """Persist audit record to aviation security compliance documentation store (stub)."""
    import json, sys
    print(json.dumps(record), file=sys.stderr)


class AdversarialAviationSecurityAIImageError(Exception):
    """Raised when an airport or aviation security AI image exceeds the adversarial injection threshold."""
    pass

Call scan_aviation_security_ai_image() with AviationSecurityAIContext.CHECKPOINT_BAGGAGE_CT_AI before forwarding Smiths Detection HI-SCAN CT-P AI, Leidos ProVision AI, or L3Harris CVR checkpoint X-ray and CT baggage images to prohibited item and explosive detection classification AI — with checkpoint_ref linking the Glyphward scan to the checkpoint lane record for 49 USC §44901 aviation security screening compliance, TSA 49 CFR Part 1544 aircraft operator security programme, and DHS OIG covert testing compliance documentation. Call with AviationSecurityAIContext.BIOMETRIC_FACIAL_RECOG_AI for CBP TVS, IDEMIA, or Clear AI biometric facial comparison and identity match display images before identity verification and Secure Flight watchlist matching AI, with screening_entity_hash as the SHA-256 of the passenger name record for 8 USC §1187 VWP identity verification, INA §275 improper entry facilitation, and TSA Secure Flight 49 CFR Part 1560 watchlist compliance audit trail. Call with AviationSecurityAIContext.TRAVEL_DOCUMENT_AUTH_AI for CBP automated passport control AI or TSA document check AI passport scan and chip verification display images before travel document authenticity and SLTD database match classification AI, with screening_entity_hash as the SHA-256 of the document number for 18 USC §1543 passport forgery bypass prevention, REAL ID Act document standards, and INA §212(a)(6)(C) misrepresentation inadmissibility compliance documentation. Call with AviationSecurityAIContext.AIR_CARGO_SCREENING_AI for CCSF or CBP NTC air cargo X-ray and CT scan images before freight threat object and contraband detection AI — with checkpoint_ref as the air waybill number for 49 CFR Part 1548 CCSF screening programme compliance, 18 USC §545 smuggling facilitation prevention, and CBP 19 USC §1595a civil penalty audit trail. Get early access

Coverage matrix

Frequently asked questions

Further reading

• Government border control AI prompt injection — related adversarial attack surface covering border security and immigration AI with CBP, DHS, and immigration law dimensions directly applicable to airport and aviation security contexts where CBP port-of-entry identity verification and travel document authentication intersect.
• CCTV and physical security AI prompt injection — related adversarial surface covering security surveillance AI with physical security, CISA, and critical infrastructure dimensions applicable to airport perimeter security and terminal CCTV monitoring AI contexts where adversarial image injection affects physical security monitoring.
• Aerospace and defence AI prompt injection — related adversarial attack surface covering aerospace and defence AI with ITAR, DoD, and national security dimensions applicable to aviation security AI contexts where defence contractor biometric and document authentication AI systems intersect with TSA and CBP aviation security programmes.
• Free tier — 10 scans/day, no card required — start scanning airport and aviation security AI images at development volumes before committing to a production plan.