Files
Evan Downing 2be6f91545 Parse result logs (#326)
* Add doc on parsing logs and scripts to unpack dataset

* Reorganize scripts

* Reorg markdown files.
2025-08-27 15:26:38 -04:00

70 lines
1.6 KiB
Markdown

# Scored
This document describes how to parse logs from the final round of the AIxCC competition.
**NOTE**: PoVs were not released to us immediately because DARPA is still involved in responsible disclosure: <https://aicyberchallenge.com/Finals-winners-announcement/>.
## Data
Download the following files from the competition to `buttercup/data/`
- `litellm_backup.sql`
- `redis-backup.rdb`
- `task-storage.tar`
## Get Task IDs for Specific Projects
For example, find the Task IDs for `libpng`.
```shell
cd data/
../scripts/results_unpack.sh
../scripts/results_find.sh libpng
```
## Parse
Query the CRS database for submissions. Documentation can be found at <https://redis.io/learn/explore/import#restore-an-rdb-file>.
Let's assume a task ID `0197b76a-b429-78fa-8d0e-ad84994c8f44` was outputted from above.
1. Start `redis-server` in `data/` directory
```shell
cd data/
redis-server
```
1. Print directory to confirm
```shell
redis-cli config get dir
1) "dir"
2) "<path/to/data>"
```
1. Stop `redis-server` and check if `dump.rdb` exists in `data/` directory.
1. Copy `.rdb` file
```shell
cp redis-backup.rdb dump.rdb
```
1. Start `redis-server` and query to make sure it has data.
```shell
redis-server
redis-cli keys \*
```
1. Parse database for task id
```shell
cd common/
uv sync --all-extras
uv run buttercup-util read_submissions --task_id 0197b76a-b429-78fa-8d0e-ad84994c8f44 --verbose True > ../data/submissions.txt
```
1. Read through `submissions.txt` for information like crash harnesses, stacktraces, and patches.