Canonical semantic map
Explore FugitiveAI’s canonical mt.* concepts, their direct relationships, and the public challenges, guides, techniques, and glossary entries that teach or demonstrate them.
How to read the graph
Every card is one canonical Concept from the registry. Its Related concepts are graph edges. Its Public resources are reviewed FugitiveAI routes where that Concept is explained or exercised.
This view exposes canonical identity, definition, safe facets, direct Concept relationships, and public resource mappings. Internal provenance, repository paths, evaluation fixtures, and durable-memory metadata are not part of the public graph.
34 stable identities
Concept IDs are preserved byte-for-byte across FugitiveAI and semantic integrations.
Core domain
Foundational identities used across hidden-channel analysis.
mt.core.steganography
The practice of concealing a secondary message inside an ordinary-looking carrier so the existence of the message is not immediately obvious.
mt.core.encoding
A reversible mapping between representations of information; encoding changes form but does not by itself hide that information exists.
mt.core.carrier
The visible or ordinary-looking artifact that contains or transports a hidden payload.
mt.core.payload
The information concealed or transmitted through a carrier.
Detection domain
Evidence, controls, interventions, and verification concepts.
mt.detection.steganalysis
The defensive analysis of artifacts for evidence that a hidden-information channel exists.
mt.detection.homologous_evaluation
An evaluation design that holds generator, topic, intent, or other process factors matched while changing the hypothesized channel or surface-language factor being tested.
mt.detection.reproducible_decoder_evidence
Evidence that a documented extraction rule repeatedly recovers the intended machine-readable information from controlled artifacts.
mt.detection.causal_channel_evidence
Evidence that deliberately destroying, randomizing, swapping, or replacing the suspected information-carrying variable causes the predicted loss or change of receiver capability while preserving relevant ordinary content.
mt.detection.false_positive_control
Methods that distinguish intentional or reproducible hidden-channel evidence from accidental patterns, confounds, multiple-testing artifacts, or benign representation differences.
mt.detection.normalization_diagnostic
The defensive use of a controlled normalization step to test which representation layer carries a suspected machine-readable state and whether that state survives transformation.
Linguistic domain
Meaning-preserving choices in language that can carry structure.
mt.linguistic.acrostic
A hidden-message construction in which selected initial letters, words, or other aligned positions form a secondary message.
mt.linguistic.word_length_parity
A structural text channel that maps odd and even alphabetic word lengths to binary states under a fixed counting rule.
mt.linguistic.sentence_length_parity
A structural text channel that maps odd and even counted sentence lengths to binary states under an explicit token-counting rule.
mt.linguistic.synonym_choice
A linguistic hidden-information method that maps curated context-appropriate lexical alternatives to deterministic states.
mt.linguistic.punctuation_choice
A text hidden-information method that maps predefined punctuation alternatives at known carrier positions to deterministic states.
mt.linguistic.contraction_choice
A linguistic method that assigns state to a curated choice between a full form and a contracted or shortened form where the language permits such alternatives.
mt.linguistic.contraction_choice.en
The English-specific realization of contraction-choice steganography using curated full forms and English contractions such as “do not” and “don’t.”
mt.linguistic.active_passive_voice
A linguistic method that maps curated active-voice and passive-voice realizations of the same underlying event to deterministic states.
mt.linguistic.sentence_order
A linguistic method that maps predefined orderings of the same textual facts or sentence units to deterministic states.
mt.linguistic.spelling_variant
A linguistic method that maps predefined context-reviewed orthographic variants to deterministic states.
mt.linguistic.spelling_variant.en
The English-specific realization of spelling-variant steganography using curated orthographic alternatives associated with English usage conventions.
Representation domain
Differences between visible, serialized, encoded, and machine-readable forms.
mt.representation.normalization
A transformation that maps multiple surface representations into a standardized representation for comparison, parsing, or processing.
mt.representation.human_visible
The content and structure ordinarily perceived from the rendered artifact without inspecting internal machine-oriented representations.
mt.representation.machine_readable
Information recoverable by software from bytes, code points, source, metadata, parsed structure, or other machine-facing representations.
mt.representation.differential
A reproducible difference between two valid representations of the same artifact, such as visual, parsed, accessibility, byte-level, or model-ingestion views.
mt.representation.unicode_code_point
The numeric identifier assigned by Unicode to a character or control function, independent of how that code point is rendered as a glyph.
mt.representation.invisible_unicode
Unicode code points present in text that may have no ordinary visible glyph while remaining machine-readable and potentially affecting parsing or layout.
mt.representation.dom
The browser-accessible tree representation produced by parsing HTML, which can differ materially from the page as visually rendered.
mt.representation.metadata
Structured information describing an artifact or its context rather than forming the artifact’s primary visible content.
mt.representation.html_comment
Source text delimited as an HTML comment that remains in source representation but is not rendered as ordinary body content.
Retrieval domain
Concept identity and semantic retrieval boundaries.
mt.retrieval.tokenizer_synchronization
The requirement that sender, receiver, or evaluator apply compatible tokenization and preprocessing so the same visible text yields compatible machine units.
No public FugitiveAI route is mapped yet.
mt.retrieval.semantic_robustness
The degree to which a machine-recognizable semantic state remains recoverable after meaning-preserving surface transformations such as paraphrase.
No public FugitiveAI route is mapped yet.
mt.retrieval.semantic_identity
The reviewed judgment that two formulations express the same underlying concept, distinguished from merely being similar or robust under transformation.
No public FugitiveAI route is mapped yet.
mt.retrieval.language_script_bias
Systematic differences in machine representation or retrieval behavior associated with language, writing system, corpus density, or tokenization rather than the underlying idea.
No public FugitiveAI route is mapped yet.