scan-vins — advanced

Synopsis

openremap scan-vins [OPTIONS] FILE

Flags

Flag	Default	Description
`--min-confidence FLOAT`	`0.4`	Only show candidates with confidence >= this value (0.0–1.0)
`--json`	—	Output as JSON instead of a table
`--help`	—	Show help

How candidates are scored

Each candidate accumulates evidence — confidence is the weighted sum, capped at 0.95. It is never a boolean claim:

Evidence	Weight
WMI in the known-manufacturer whitelist (`WVW`, `WAU`, `WBA`, …)	+0.30
ISO 3779 position-9 check digit valid	+0.25
Model-year character plausible (position 10)	+0.10
Positions 12–17 all digits	+0.10
Candidate inside an ident block (layout segmenter)	+0.10
Mirror consensus — same VIN appears multiple times	+0.10

Candidates with fewer than 6 distinct characters (fills like 99999999999999999) are rejected up front.

JSON output

[
  {
    "offset": 8192,
    "vin": "WVWZZZ1JZXW000001",
    "confidence": 0.9,
    "wmi_known": true,
    "check_digit_ok": true,
    "year_plausible": true,
    "numeric_tail": true,
    "in_ident_block": true,
    "mirror_count": 2
  }
]

Examples

# Find everything (default threshold)
openremap scan-vins stock.bin

# Only strong candidates — for a cloning/merge audit
openremap scan-vins stock.bin --min-confidence 0.6 --json

Notes

Real VINs live in ident blocks and are usually mirrored; natural lookalikes (part numbers, serials) score ≤ 0.4 on the measured corpus.
Two distinct high-confidence VINs in one file is the classic sign of a cloned or merged dump — see also openremap health.
Every candidate is decoded with vininfo (BSD-3): the JSON carries manufacturer, region, country, years, checksum_valid, and decoded; the table appends a dim — Make, Country, Year (decoded, unverified) suffix. Decoding is permissive — unknown WMIs yield decoded: false with no guesswork, and malformed input never errors.