Collected molecules will appear here. Add from search or explore.
A research code repository for benchmarking the robustness of machine learning models against adversarial attacks specifically on COVID-19 genomic sequences.
Defensibility
stars
3
forks
1
The project is an academic artifact tied to a specific paper published during the pandemic. With only 3 stars and no activity in nearly four years, it lacks any form of community traction or software ecosystem. From a competitive standpoint, the repository serves more as a 'reproducibility package' than a living tool. The defensibility is near-zero as the techniques (adversarial perturbations on strings/sequences) are well-documented in general ML security literature and can be reimplemented by any researcher in the field. Frontier labs have no interest in this specific niche, but the project is effectively displaced by broader, more robust adversarial toolkits like IBM's Adversarial Robustness Toolbox (ART) or more advanced foundation models for biology (like Nucleotide Transformer or ESM) which have superseded the simple classification models this project benchmarks. It is of historical academic interest rather than commercial or strategic value.
TECH STACK
INTEGRATION
reference_implementation
READINESS