{
  "schema": "signal-integrity-research-manifest/1",
  "release": "v2026.08.16.13",
  "authority": "operator_supplied_research_inputs",
  "truth_boundary": "These files are preserved research inputs. Their presence and SHA-256 values prove local byte preservation only; they do not constitute external verification, operational authority, or permission to create covert channels.",
  "raw_preservation": "byte_for_byte; originals are not normalized or rewritten",
  "file_count": 17,
  "files": [
    {
      "filename": "AI Image Metadata Processing Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/AI Image Metadata Processing Research.md",
      "sha256": "7961d4d87db7686748e12319df12e68ba48404c84ac596e6026055c83b1cd82f",
      "size_bytes": 35680,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**Security Implications of Accessibility Metadata in Multimodal AI Systems: Processing Mechanisms and Injection Vulnerabilities**"
    },
    {
      "filename": "Controlled, Ethical Methodology for Studying Hidden or Machine-Readable Messages Intended for AI Systems.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Controlled, Ethical Methodology for Studying Hidden or Machine-Readable Messages Intended for AI Systems.md",
      "sha256": "e3a7ddfa68048a9bfca31c21b1c0f5ba3b85116c55e640418cf8d812491450e6",
      "size_bytes": 51069,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Controlled, Ethical Methodology for Studying Hidden or Machine-Readable Messages Intended for AI Systems"
    },
    {
      "filename": "Defensive Preprocessing for Hidden Instructions in Documents and Webpages Sent to AI Systems.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Defensive Preprocessing for Hidden Instructions in Documents and Webpages Sent to AI Systems.md",
      "sha256": "b984d635d186e3834f12db8a4cb6e47dcfd755f0b90b926ca77d475ac5bf37bf",
      "size_bytes": 68956,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Defensive Preprocessing for Hidden Instructions in Documents and Webpages Sent to AI Systems"
    },
    {
      "filename": "Document Metadata in AI Processing.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Document Metadata in AI Processing.md",
      "sha256": "b17aa2c29a34e2fbf70334d12061025856a726e592282df8ddd59acdd213dbc6",
      "size_bytes": 48133,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Invisible Attack Surface: Document Metadata Exploitation in AI Retrieval and Processing Systems**"
    },
    {
      "filename": "Hidden HTML Text Defensive Analysis.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Hidden HTML Text Defensive Analysis.md",
      "sha256": "2a9a04dbad0c2b7bb3495cbad1b8963c5bd4ed66fb886a456e6839c589b35c6d",
      "size_bytes": 46412,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**Defensive Assessment of HTML Obfuscation and Hidden Text Mechanisms in Modern Web Architectures**"
    },
    {
      "filename": "Hidden Information in Natural Language Through Synonym and Word-Choice Encoding.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Hidden Information in Natural Language Through Synonym and Word-Choice Encoding.md",
      "sha256": "e736ea7f8406d8dca916829ec1a656f737b794d5df661f725a94c652b0e17d54",
      "size_bytes": 54037,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Hidden Information in Natural Language Through Synonym and Word-Choice Encoding"
    },
    {
      "filename": "Indirect Prompt Injection Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Indirect Prompt Injection Research.md",
      "sha256": "5e29dca0ec326c936fe12af8cf495a96cc871cfecdbbf44e4e15b0302076cfd5",
      "size_bytes": 42215,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Anatomy of Indirect Prompt Injection: Threat Models, Attack Mechanics, and Modern Defenses in LLM-Integrated Agents**"
    },
    {
      "filename": "Invisible Unicode Security And Processing.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Invisible Unicode Security And Processing.md",
      "sha256": "0553db9ef963c84b1db7f1bfe8f5ed85d32f7b861a9420060917963f57e3e6aa",
      "size_bytes": 45051,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Invisible Attack Surface: A Comprehensive Analysis of Zero-Width and Bidirectional Unicode Vulnerabilities**"
    },
    {
      "filename": "Invisible Unicode and Modern Language-Model Tokenizers.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Invisible Unicode and Modern Language-Model Tokenizers.md",
      "sha256": "bab924fd392ecaf6eea68ab257b64f53dce901a69e947d6819b69abc6737728e",
      "size_bytes": 54616,
      "nfc": false,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Invisible Unicode and Modern Language-Model Tokenizers"
    },
    {
      "filename": "LLM Encoding Interpretation Security.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/LLM Encoding Interpretation Security.md",
      "sha256": "92d9068d1b36fa85c31859cc6e3a1c92311aae2fb573a31d8435dd8ae8b071eb",
      "size_bytes": 59877,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Interpretation of Encoded Data by Large Language Models: Auditing, Security, and Detection Implications**"
    },
    {
      "filename": "LLM Text Steganography Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/LLM Text Steganography Research.md",
      "sha256": "a73cde41b452f09276569dfd3192225b4501f5c87887cfaf87078c1cb370d870",
      "size_bytes": 55726,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**Machine-Targeted Linguistic Signals: Steganography, Watermarking, and Covert Communication in Large Language Models**"
    },
    {
      "filename": "Linguistic Steganography Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Linguistic Steganography Research.md",
      "sha256": "f5d0fdcc4948374d61768f083b5b46f67744a4194f6bdd69509c16ea5e20055a",
      "size_bytes": 71451,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Encoding of Covert Information Through Lexical Substitution and Generative Language Models**"
    },
    {
      "filename": "Semantic Text Steganography Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Semantic Text Steganography Research.md",
      "sha256": "ba32cee5414ad93adbedc99577efe488920aa34966c484e1a26ea9df1b9c628c",
      "size_bytes": 112957,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**Advanced Frameworks in Semantic Steganography: Vocabulary-Class Encoding, Generative Models, and Information-Theoretic Security**"
    },
    {
      "filename": "Semantic-Category and Language-Model Text Steganography.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Semantic-Category and Language-Model Text Steganography.md",
      "sha256": "5d1530c1a569e5f72149433e3bd79e02e1ba76772426e7afc1ab88f871ac8110",
      "size_bytes": 72960,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Semantic-Category and Language-Model Text Steganography"
    },
    {
      "filename": "Structural Steganography in Normal-Looking English Writing.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Structural Steganography in Normal-Looking English Writing.md",
      "sha256": "20151fe63a7660a856ddefda1a2c3a4c88a42b685d1a6a40b2681beb0d216dd8",
      "size_bytes": 52106,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "Structural Steganography in Normal-Looking English Writing"
    },
    {
      "filename": "Text Steganography Techniques Analysis.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Text Steganography Techniques Analysis.md",
      "sha256": "e8d9983d48cd3fb9f6b7407948db710f0ff8d54a41955bda2df92b1435c4e09f",
      "size_bytes": 50244,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**Structural and Linguistic Text Steganography: Advanced Encoding Mechanisms, Robustness, and Steganalysis**"
    },
    {
      "filename": "Unicode Homoglyph Attacks Research.md",
      "path": "docs/source-evidence/operator-uploaded/2026-08-16-signal-integrity/Unicode Homoglyph Attacks Research.md",
      "sha256": "40f19adcdaf51b7badc472e9c03ba87843622e80242ea36d921dceb8bb1b9b25",
      "size_bytes": 48923,
      "nfc": true,
      "unexpected_control_count": 0,
      "unexpected_controls": [],
      "title": "**The Invisible Vector: Analyzing Unicode Homoglyph and Look-Alike Character Attacks Across Digital Infrastructure**"
    }
  ]
}