# Break it on the bench, not in the field.

> The fuzzing engine sends randomised, reproducible traffic to an ECU in two modes. UDS fuzzing sends random payloads (up to 128 bytes) on selected or non-standard service IDs over ISO-TP, re-entering the diagnostic session periodically. CAN fuzzing sends random CAN and CAN FD frames on arbitration IDs taken from a DBC. Runs are seeded, so a finding can be reproduced exactly. Crash detection runs after every iteration: a TesterPresent liveness check (always on), an optional DTC-count monitor, and timeout or transport-error detection. Progress streams live and results export to PDF.

AutoST fuzzes ECUs over UDS and CAN with reproducible seeds and live crash detection via TesterPresent and DTC monitoring. Built for benches, never for the road.

Source: https://auto-st.com/ecu-fuzzing

AutoST throws malformed and unexpected traffic at an ECU and watches for the moment it stops responding, resets or throws a fault. Every run is reproducible, so a crash can be replayed and handed to developers.

## UDS and CAN, your choice

### UDS fuzzing

Random payloads on standard or non-standard service IDs, with session re-entry, configurable timeouts and an exclusion list.

### CAN fuzzing

Random classic and FD frames on arbitration IDs from your DBC, with a reproduce count to confirm a finding before it is logged.

## How AutoST knows something broke

### Liveness

A TesterPresent check after every iteration catches an ECU that has stopped responding. Always on.

### DTC monitor

Optionally watches the diagnostic trouble code count and flags any change during the run.

### Timeouts

Timeouts and transport errors on a fuzzed request are recorded as findings, with the exact payload.

## A crash you can replay

Every fuzzing run is driven by a seed that AutoST stores and shows. Re-run with the same seed and you get the same sequence, so a crash is not a one-off you can never find again.

Each finding is recorded with its iteration, timestamp, the monitor that caught it and the payload, plus a short history of the last payloads leading up to it.

- Seeded, reproducible runs
- Payload and monitor recorded per finding
- Live iterations, crashes and requests per second
- Exclusion lists for IDs and service IDs

## For the bench only

Fuzzing can put an ECU into a fault state, reset it or brick it. AutoST is a bench tool: never fuzz a vehicle in traffic or a system whose failure could hurt someone.

- Run on an isolated bench
- Never on public roads
- Keep a recovery path for the ECU

## FAQ

**Is the fuzzing reproducible?**

Yes. Each run uses a stored seed, so you can replay the exact sequence that triggered a crash and hand it to developers.

**How does it detect a crash?**

With a TesterPresent liveness check after every iteration, an optional DTC-count monitor, and timeout or transport-error detection on the fuzzed requests.

**Can I fuzz a car on the road?**

No. Fuzzing is a bench activity. It can put an ECU into a fault state, so it must never run on a vehicle in traffic.

---
AutoST by Zyberum. Canonical page: https://auto-st.com/ecu-fuzzing
