scanNormalization()

Defined in: normalization.ts:40

Unicode normalization detector.

Detects characters that change under NFKC normalization.

Unicode normalization can transform visually similar characters into canonical equivalents. When user-visible text differs from its normalized form, this may indicate:

• spoofing attempts
• homoglyph confusion
• prompt-smuggling techniques
• content-validation bypass

Detection model:

• Compare each character against its NFKC-normalized form
• Group adjacent normalization-sensitive characters into a span
• Emit one threat per span

NOTE: This is intentionally heuristic. Many normalization changes are legitimate in multilingual text, but normalization-sensitive spans in prompts or code should be surfaced for inspection.

Span semantics: offendingText = original span decodedPayload = normalized span

Rule: PSN001 — Normalization-sensitive text detected

Parameters

text

string

options?

ScanOptions = {}

context?

ScanContext = {}

Returns

ThreatReport[]