{
  "id": "dai-cai-2019",
  "title": "Towards Near-imperceptible Steganographic Text",
  "authors": [
    "Falcon Z. Dai",
    "Zheng Cai"
  ],
  "author_parts": [
    {
      "given": "Falcon Z.",
      "family": "Dai"
    },
    {
      "given": "Zheng",
      "family": "Cai"
    }
  ],
  "year": 2019,
  "venue_date": "2019-07",
  "venue": "Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics",
  "venue_short": "ACL 2019",
  "publication_type": "conference",
  "publisher": "Association for Computational Linguistics",
  "pages": "4303–4308",
  "doi": "10.18653/v1/P19-1422",
  "arxiv_id": "1907.06679",
  "relation": "related",
  "categories": [
    "Contemporary 2019 research",
    "Statistical security",
    "Huffman coding"
  ],
  "summary": "Analyzes how binning and variable-length encoding alter a language model's output distribution. Its patient-Huffman method skips message embedding at steps where the induced distribution would differ too much from the model.",
  "relationship": "Neural Linguistic Steganography explicitly identifies this as concurrent research on distributional analysis and modified Huffman coding, distinct from its arithmetic-coding approach.",
  "limitation": "Tight total-variation bounds require restrictive per-step thresholds, with useful throughput depending on the model's distributions; the paper also identifies vulnerability to an active adversary modifying text.",
  "sources": [
    {
      "url": "https://aclanthology.org/P19-1422/",
      "evidence": "Published title, authors, venue, date, DOI, pages."
    },
    {
      "url": "https://arxiv.org/abs/1907.06679",
      "evidence": "First arXiv submission July 15, 2019."
    },
    {
      "url": "https://aclanthology.org/P19-1422.pdf",
      "evidence": "Section 3.3 gives patient-Huffman and threshold qualification; section 4 identifies active-adversary limitations."
    },
    {
      "url": "https://aclanthology.org/D19-1115.pdf",
      "evidence": "Section 2 calls Dai and Cai concurrent work."
    },
    {
      "url": "https://github.com/falcondai/lm-steganography",
      "evidence": "Author repository explicitly accompanies the paper and implements patient-Huffman."
    }
  ],
  "arxiv_url": "https://arxiv.org/abs/1907.06679",
  "earliest_date": "2019-07-15",
  "earliest_date_label": "First arXiv submission",
  "url": "https://aclanthology.org/P19-1422/",
  "pdf": "https://aclanthology.org/P19-1422.pdf",
  "code": "https://github.com/falcondai/lm-steganography",
  "category": "security",
  "contribution": "Analyzes how binning and variable-length encoding alter a language model's output distribution. Its patient-Huffman method skips message embedding at steps where the induced distribution would differ too much from the model.",
  "canonical_url": "https://steganography.live/reading-list/papers/dai-cai-2019/"
}