code/attack_generation/README.md

# IV-CAN-v1 attack-generation snapshot

This directory is the self-contained reproducibility snapshot of the attack
construction code used by IV-CAN-v1. It preserves the dataset taxonomy and
frame-generation logic while excluding private builder infrastructure and the
third-party `PCANBasic.dll` binary.

## Implemented scope

| Attack family | Dry-run plan | Hardware execution |
| --- | --- | --- |
| DoS | Yes | Side-channel injection |
| fuzzing | Yes | Side-channel injection |
| replay | Yes | Side-channel injection |
| spoofing | Yes | Side-channel injection |
| suspension | Yes | Not implemented; requires a two-interface MitM bridge |
| masquerade | Yes | Not implemented; requires a two-interface MitM bridge |

The values accepted by each family match `label.yaml` and are documented in
[`../../SCHEMA.md`](../../SCHEMA.md).

## Install

Dry-run mode uses only the Python standard library:

```bash
python -m pip install -e .
```

Hardware use requires the optional PCAN adapter dependency plus the official
PEAK-System driver/runtime installed under its own license:

```bash
python -m pip install -e '.[pcan]'
```

No PEAK-System DLL is redistributed in this repository.

## Safe default: preview only

The CLI does not transmit unless explicitly requested:

```bash
ivcan-attack --type DoS \
  --id-option zeroId \
  --dlc-option dlcChange \
  --frame-format standardId \
  --data-option FFData \
  --duration 10
```

Replay can consume an IV-CAN-v1 CSV or the synthetic example:

```bash
ivcan-attack --type replay \
  --timing immediate \
  --replay-file examples/synthetic_replay.csv
```

Hardware transmission requires both `--execute` and
`--acknowledge-bench-only`. It must only be used on an authorized, isolated
test bench. Never connect this tool to a vehicle operating on a public road.

## Reproducibility notes

- `timestamp` differences are used to reproduce replay timing.
- `is_extended_id` and `is_fd` are consumed when present in replay CSV files.
- The `seed` option makes generated DoS/fuzzing previews deterministic.
- Suspension and masquerade are published as validated experiment plans, not
  as claims of an implemented MitM bridge.

## Tests

```bash
python -m unittest discover -s tests -v
```

This snapshot is distributed under CC BY 4.0 with the rest of IV-CAN-v1.
Third-party hardware drivers and runtimes remain under their vendors' terms.