Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension


Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 2 additions & 2 deletions Cargo.lock

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

12 changes: 7 additions & 5 deletions SECURITY.md
Original file line number Diff line number Diff line change
Expand Up @@ -22,11 +22,13 @@ it, and what it doesn't protect against:
The `train` tool writes the weights it trains to a `weights` folder in the folder it's
run from, `bench` writes no files, and the app's screenshot mode (`TS_SHOT`) writes
only the image you name.
* **Files are checked before use.** A weights file is parsed into typed data and checked
against what its controller accepts (layer sizes, connections between layers, finite
values). A file that fails is rejected rather than run, and CI fuzzes this loader on
every pull request. A file that passes can still make its controller manage traffic
badly, so use weights only from sources you trust.
* **Files are checked before use.** A weights file is parsed into typed data, and its
network is checked against what its controller accepts (layer sizes, connections
between layers, finite values). A file that doesn't parse is skipped for the next
copy. A network that fails the check is never run: a Deep RL file with one is skipped
too, and a Hybrid file loads without its network and runs only its saved base tuning.
CI fuzzes this loader on every pull request. A file that loads can still make its
controller manage traffic badly, so use weights only from sources you trust.
* **No code from data.** The app has no plugins, scripting or other way to run code
from the files it reads, and its own code is safe Rust with no `unsafe` blocks.
* **Verifiable releases.** Every release file is signed and comes with a
Expand Down
2 changes: 1 addition & 1 deletion THIRD_PARTY_LICENSES.md
Original file line number Diff line number Diff line change
Expand Up @@ -5228,7 +5228,7 @@ Used by:

* adler2 2.0.1 (https://github.com/oyvindln/adler2)
* async-channel 2.5.0 (https://github.com/smol-rs/async-channel)
* async-executor 1.14.0 (https://github.com/smol-rs/async-executor)
* async-executor 1.14.1 (https://github.com/smol-rs/async-executor)
* async-io 2.6.0 (https://github.com/smol-rs/async-io)
* async-lock 3.4.2 (https://github.com/smol-rs/async-lock)
* async-process 2.5.0 (https://github.com/smol-rs/async-process)
Expand Down
4 changes: 2 additions & 2 deletions docs/assurance-case.md
Original file line number Diff line number Diff line change
Expand Up @@ -114,8 +114,8 @@ error, also check all code on every pull request.
* **One maintainer.** The maintainer's own changes are checked by CI, CodeQL and fuzzing
rather than by a second reviewer, and the release pipeline trusts the maintainer's
GitHub account: whoever controlled that account could publish a release.
* **Valid but unhelpful weights.** A weights file that passes every check can still make
its controller manage traffic badly. That affects the simulation's results, not the
* **Valid but unhelpful weights.** A weights file that loads can still make its
controller manage traffic badly. That affects the simulation's results, not the
computer it runs on.
* **Very large files.** A weights file is read whole into memory before it's checked, so
an enormous one can make the app slow or exhaust its memory.
Expand Down
Loading