Validation¶

Structural validation and the deterministic confidence heuristic.

Confidence is not a probability and it is not a model score. It is a fixed function of the type, the validation result, and noise tags. The same indicator always receives the same score.

Functions:

ipv4_class –

Classify an IPv4 address the way the sieve tags noise.
ipv6_tags –

Return a single noise tag for an IPv6 address, or an empty tuple.
validation_notes –

Return reason codes for a normalized value. Empty means structurally valid.
domain_notes –

Validate a host name against label rules and the bundled TLD list.
confidence_for –

Score an indicator from its validation result and noise tags.

ipv4_class ¶

ipv4_class(ip: str) -> str | None

Classify an IPv4 address the way the sieve tags noise.

Parameters:

ip (str) –

Dotted IPv4 address. Octets are not re-validated here.

Returns:

str | None –

private, loopback, link-local, cgnat, multicast,
str | None –

reserved, or None for an ordinary public address.

ipv6_tags ¶

ipv6_tags(address: IPv6Address) -> tuple[str, ...]

Return a single noise tag for an IPv6 address, or an empty tuple.

Tags come from :mod:ipaddress: loopback, link-local, multicast, private (unique-local), then reserved or unspecified.

validation_notes ¶

validation_notes(kind: str, normalized: str) -> tuple[str, ...]

Return reason codes for a normalized value. Empty means structurally valid.

Parameters:

kind (str) –

Type key.
normalized (str) –

Canonical value.

Returns:

tuple[str, ...] –

A tuple of stable codes such as tld-not-recognized.

domain_notes ¶

domain_notes(domain: str) -> tuple[str, ...]

Validate a host name against label rules and the bundled TLD list.

Parameters:

domain (str) –

Normalized host name.

Returns:

str –

Reason codes. tld-not-recognized means the last label is neither in
... –

the bundled root-zone list nor in the small special-use set.

confidence_for ¶

confidence_for(*, kind: str, valid: bool | None, tags: tuple[str, ...]) -> float

Score an indicator from its validation result and noise tags.

The result is rounded to two decimals and clamped to 0.01-0.99.

Invalid: 0.25
Validation skipped: 0.55
Valid: 0.80
Tagged IP, IPv6, or CIDR: subtract 0.15
Valid untagged IP or IPv6: add 0.10
Valid hash, JA3, JA3S, or JARM: add 0.10

Parameters:

kind (str) –

Type key.
valid (bool | None) –

Validation result, or None when validation did not run.
tags (tuple[str, ...]) –

Noise tags.

Returns:

float –

A heuristic score. It is not a probability.