EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Sep 9·5 min read
Factoring RSA 260
TL;DR
A new record in RSA factoring was achieved by the Cognition research team, who developed a high-performance GPU lattice siever to factor RSA-260 at a significantly reduced cost. This breakthrough demonstrates the potential for hyperscalers to factor RSA-1024 numbers more economically, though RSA-2048 remains largely unaffected.
✦ Why It Matters
Evaluate your cryptographic strategies, especially if relying on RSA-2048 for security.
Key Takeaways
- Consider using GPU acceleration for computational number theory tasks.
- Investigate the general number field sieve for large number factorizations.
- Optimize job scheduling to utilize fragmented compute resources effectively.
- Stay informed about advancements in cryptanalysis to assess security risks.
How It Works
The factorization of RSA-260 was achieved using a modified implementation of the general number field sieve (GNFS) optimized for GPUs. The process involved several stages: polynomial selection, lattice sieving, and linear system solving.
The lattice sieving stage, which is computationally intensive, was adapted to leverage the high memory bandwidth of GPUs, allowing for significant performance improvements over traditional CPU methods.