GenoAdv

Ethical Use and Dual-Use Disclosure

Important Notice on Dual-Use and Misuse Risks

The GenoAdv dataset contains adversarial genomic sequences designed to evaluate the robustness of genomic foundation models (GFMs), particularly those used in tasks such as gene pathogenicity prediction and clinical diagnostics. While this dataset was developed for research purposes aimed at improving the security and reliability of genomic models, it also exposes critical vulnerabilities that could be misused if applied irresponsibly.

We acknowledge that adversarial attacks targeting biologically meaningful regions may pose risks in sensitive domains. Potential misuse scenarios include:

• Attempting to evade genomic screening or diagnostic systems,
• Manipulating clinical model outputs to mislead healthcare decisions,
• Interfering with biosecurity screening pipelines.

To mitigate such risks, GenoAdv is released under a research-only, non-commercial license. Use of this dataset is strictly limited to academic or non-commercial research activities. Any use in clinical, diagnostic, or safety-critical applications is expressly prohibited. Redistribution or deployment of this dataset or its derivatives for harmful purposes is a violation of ethical AI research standards.

By downloading or using this dataset, you agree to:

• Use it only for responsible, non-commercial, research purposes,
• Not employ it in any real-world diagnostic or medical decision-making systems,
• Refrain from attempts to exploit model vulnerabilities in deployed systems.

If you have questions about responsible use or wish to apply for access under special conditions (e.g., for use in authorized robustness evaluations), please contact the authors.

We believe that openly identifying and studying these vulnerabilities is essential for building safer and more trustworthy AI in genomics. This dataset is provided in the spirit of proactive defense—not offense.

How to download this dataset

git lfs install

git clone https://huggingface.co/datasets/magicslabnu/GFM-Attack

How to use data

Each directory corresponds to a dataset and contains the standard files: train.csv, dev.csv, and test.csv. You may select any of these dataset folders to perform adversarial training. For example, to use the tf1 dataset for adversarial training, utilize the train.csv file located within the tf1 folder.

Paper

https://arxiv.org/abs/2505.10983