RSA-896
summary
Stephen A. Weis used Claude to port CADO-NFS to GPUs, allowing for the factorization of the RSA-896 challenge number. By leveraging a fleet of 2,048 GPUs during idle time, the project achieved 30 GPU-years of compute power over ten days. This achievement demonstrates that RSA-1024 keys are now vulnerable to actors with data center-level GPU resources. While the General Number Field Sieve algorithm itself was not improved, the scale of available compute makes smaller RSA keys insecure.
key points
- Claude acted as an orchestrator to port CADO-NFS to GPU acceleration and manage a massive fleet of hardware across Anthropic's infrastructure.
- The factorization was achieved using 30 GPU-years of compute time, showing that brute-force scale outweighs algorithmic breakthroughs in this instance.
- The result proves that RSA-1024 is no longer secure against organizations with access to large-scale GPU clusters.
who should read it
Cybersecurity professionals and cryptographers interested in the RSA challenge and the practical application of AI in high-performance computing.
discussion positive
consensus
Commenters agree that while the use of AI to port CADO-NFS to GPUs is an interesting application of LLMs, the achievement is primarily a result of massive compute power rather than algorithmic breakthroughs. There is a general understanding that RSA-1024 remains practically secure for most, though nation-states with millions of GPUs could potentially crack it. The discussion also emphasizes that using 'scavenged idle capacity' in pre-paid data centers makes such projects financially feasible by eliminating marginal costs.
top insight
A detailed resource calculation suggests that factoring RSA-1024 would require approximately 585 GPU-years, costing roughly $20 million using similar setups. This provides a concrete scale for the remaining security margin of 1024-bit keys, moving the conversation from theoretical risk to a specific financial and temporal estimate.
Generated by an LLM from the article and thread. Check the sources before relying on it.