`openremap scan-vins`

Locate VIN candidates in an ECU binary and score them on structural evidence — never a bare claim: WMI whitelist, ISO 3779 check digit, model-year character, numeric tail, ident-block context, mirror consensus. Every candidate is also decoded (vininfo): WMI → manufacturer, region, country, and model years — labelled decoded, unverified.

New here? Read the plain-English introduction first.

The API name for this method is scan_vins (underscore) — see scan-vins — API.

Synopsis

openremap scan-vins <FILE> [OPTIONS]

Argument	Required	Description
`FILE`	Yes	ECU binary to scan (`.bin` / `.ori` / `.hex`). The file must exist.

Options

Option	Default	Description
`--min-confidence FLOAT`	`0.4`	Only show candidates with confidence ≥ this value (0.0–1.0).
`--json`	off	Output the full result as JSON instead of a table.
`--help`		Show help and exit.

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

Ident-block and mirror evidence only corroborate candidates that already look like a VIN (a known WMI) — serials and calibration IDs are mirrored too, and without the gate they would cross the lookalike line. Candidates with fewer than 6 distinct characters (fills like 99999999999999999) are rejected up front.

Measured on the real corpus: natural lookalikes score ≤ 0.45; injected real-shaped VINs in ident blocks score ≥ 0.9.

Example

# 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

# Full JSON output
openremap scan-vins stock.bin --json

Example output

  OpenRemap — VIN Scan
  stock.bin  •  2,097,152 bytes  •  1 candidate(s) ≥ 0.6

   Offset       VIN                 Conf  Evidence
  ──────────────────────────────────────────────────────────────────
  0x002000  WVWZZZ1KZ7W059972       0.90  wmi, check-digit, year, numeric-tail, ident-block, mirrored-x2  — Volkswagen, Germany, 2007 (decoded, unverified)

When no candidate reaches the cut-off, the command prints No VIN candidates found at this confidence level.

JSON output

{
  "file": "stock.bin",
  "file_size": 2097152,
  "candidates": [
    {
      "offset": 8192,
      "vin": "WVWZZZ1KZ7W059972",
      "confidence": 0.9,
      "evidence": ["wmi", "check-digit", "year", "numeric-tail", "ident-block", "mirrored-x2"],
      "mirror_count": 2,
      "manufacturer": "Volkswagen",
      "region": "Europe",
      "country": "Germany",
      "years": [2007],
      "checksum_valid": false,
      "decoded": true
    }
  ]
}

(Fields above follow the real schema; the specific values are from an example run. The JSON report is exactly the object the API method returns.)

Notes

Real VINs live in ident blocks and are usually mirrored; natural lookalikes (part numbers, serials) score ≤ 0.45 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.
scan-vins never modifies the file — it only reads it.

openremap scan-vins