Skip to content

Latest commit

 

History

History
65 lines (42 loc) · 1.9 KB

File metadata and controls

65 lines (42 loc) · 1.9 KB

FunctionFoundry.Text

Complex text analysis for spoof detection, secret hunting, near-duplicates, and CSV/TSV dialect discovery.

What this product does

It analyzes text for risk signals and similarity, with explainable results and hard processing limits.

1. Unicode spoof / confusable detection (UnicodeSpoofDetector)

Finds lookalike and mixed-script abuse patterns:

  • Unicode normalization checks
  • confusable skeleton generation (versioned compact dataset)
  • mixed-script detection
  • invisible / control-character findings
  • explainable risk findings

Does not claim that a heuristic result proves malicious intent.

2. Secret-candidate scanner (SecretCandidateScanner)

Scans large text for likely secrets:

  • high-entropy tokens
  • configurable patterns
  • context-aware confidence scores and false-positive controls
  • streaming support with line/column reporting
  • redacted findings by default

3. Near-duplicate text index (NearDuplicateTextIndex)

Indexes documents with MinHash / LSH-style techniques:

  • configurable similarity thresholds
  • incremental indexing
  • deterministic seeds
  • candidate retrieval plus exact similarity verification
  • documented memory / accuracy trade-offs

4. Delimited-text dialect inference (DelimitedTextDialectInferrer)

Infers delimiter, quote, escape, newline, and header likelihood from samples. Returns ranked candidates with confidence and evidence — never silently claims certainty. Sample size and complexity are limited.

When to use it

  • Homograph / phishing username checks
  • Pre-commit or CI secret scanning of text blobs
  • Deduplicating support tickets, docs, or logs
  • Auto-detecting messy CSV dialects before parsing

Non-goals

  • Full OCR or NLP pipelines
  • Guaranteeing zero secret false positives

Install

dotnet add package FunctionFoundry.Text

Runtime dependencies

None beyond the .NET shared framework.