# FAQ

### **Q: What datasets are provided?**

The dataset includes:

* 500K profiling/training traces (random plaintext and key);
* 100K attack traces (random plaintext, fixed key).

All traces are stored in an HDF5 file, similar to ASCAD’s structure, with metadata (plaintext, key, labels).\
See Datasets

### Q: What is the evaluation metric?

Submissions are ranked by average $$ge\_{+ntge}$$ (see [Challenge Rule](https://pace-tl.gitbook.io/ches-challenge-2025/quickstart-4)) over four attack trace sets (1 public, 3 private). Lower $$ge\_{+ntge}$$ represents better performance.

### Q: Are there restrictions on libraries or frameworks?

Yes. Code must use PyTorch 2.7.0 or TensorFlow 2.19.0 and all the libraries given in the requirements.txt file (see [Getting Started](https://pace-tl.gitbook.io/ches-challenge-2025/getting-started)). Submissions with incompatible versions will be disqualified.

### Q: Can I submit multiple times?

Yes! Participants can make multiple submissions during the challenge period (from June 15 to August 15).  Each submission is limited to one attack.

### Q: How are submissions validated?

Organizers test submissions against:

* The public attack trace set;
* Three private attack trace sets;
* Correctness of key recovery.

Invalid submissions (e.g., rule violations) are discarded.

### Q: What should a submission include?

Submission should include,

* ReadMe filed describing the GE and NTGE;
* Codebase;

See [Submission](https://pace-tl.gitbook.io/ches-challenge-2025/quickstart-3) for more detailed explanations.

### Q: How is the scoreboard updated?

The scoreboard is updated continuously. Final rankings will be announced around September 1.