Skip to content

Agent Output Steganography Detector -- finds hidden messages in agent text outputs

Agent Output Steganography Detector — finds hidden messages in agent text outputs.

Module: replication.steganography

Quick Start

from replication.steganography import StegoVector

instance = StegoVector()

Enums

StegoVector

  • WHITESPACE
  • ACROSTIC
  • HOMOGLYPH
  • CAPITALIZATION
  • PUNCTUATION
  • SENTENCE_LENGTH
  • INVISIBLE_UNICODE
  • SYNONYM_SUBSTITUTION

RiskLevel

  • NONE
  • LOW
  • MEDIUM
  • HIGH
  • CRITICAL

Classes

StegoVector

RiskLevel

StegoFinding

StegoReport

Method Description
has_findings()
findings_by_vector()
highest_risk()
summary()

StegoConfig

SteganographyDetector

Analyses agent text output for steganographic encodings.

Method Description
__init__()
analyze()
batch_analyze()
compare_texts()
detect_whitespace_encoding()
detect_invisible_unicode()
detect_acrostic()
detect_homoglyphs()
detect_capitalization_encoding()
detect_punctuation_anomalies()
detect_sentence_length_encoding()

Functions

Function Description
scan_text()
encode_zero_width() Encode a message using zero-width characters (for testing).
encode_homoglyphs() Encode bits via homoglyph substitution (for testing).