Invisible text scanner

cerbia.core.scanners.InvisibleTextScanner detects Unicode characters that are easy for a person to miss but can change how downstream text is interpreted. It counts zero-width characters, bidi overrides, format controls, tag characters, and private-use or unassigned code points.

Parameter Default Meaning

Parameter	Default	Meaning
`threshold`	`3`	Number of detected invisible characters allowed before high-risk handling.
`languages`	all registered	Language packs used for suspicious-keyword context after characters are removed.
`severity`	`CRITICAL`	Finding severity.
`action`	`BLOCK`	Finding action.
`content_types`	`null`	Accepted content types; `null` accepts every type.

threshold

3

Number of detected invisible characters allowed before high-risk handling.

languages

all registered

Language packs used for suspicious-keyword context after characters are removed.

severity

CRITICAL

Finding severity.

action

BLOCK

Finding action.

content_types

null

Accepted content types; null accepts every type.

No hidden characters yields risk 0.0. Counts at or below threshold produce risk 0.2; larger counts begin at 0.7 + count × 0.03, capped at 1.0. If the cleaned text matches a suspicious keyword pattern, the score receives an additional 0.15 bonus.

scanners:
  - scanner: cerbia.core.scanners.InvisibleTextScanner
    init_args:
      threshold: 3
      languages: ["en", "es"]