MiniMax Uses Latest Internal Model to Break Multiple Frontier Cryptographic Schemes, Setting New Records in International Challenges
IT Home, November 22 — MiniMax announced today that its research team has recently used MiniMax's latest internal model to conduct systematic automated cryptanalysis of modern cryptographic schemes.
In testing, the model uncovered multiple previously unknown algorithmic construction flaws and advanced several attacks that had previously remained at the theoretical paper level to executable breaks achievable in seconds to minutes.
IT Home has learned that the relevant progress mainly covers three areas:
- For past NIST (National Institute of Standards and Technology) international cryptographic standardization competitions, the internal model for the first time actually broke multiple candidate algorithms that previously had "no known practical attack";
- For papers and competition-winning schemes published in recent years at top international cryptography conferences and journals, the internal model achieved efficient ciphertext forgery or key recovery against multiple frontier schemes that had undergone peer review, and set a new record for the highest number of broken rounds of a new type of block cipher;
- In solving classic mathematical hard problems that serve as the cornerstone of public-key cryptosystems, the internal model demonstrated the ability to combine "mining implicit algebraic structures" with "system-level engineering optimization," and in an international public challenge for the first time conquered dozens of high-order challenge instances with no publicly known solutions, achieving measured speedups of hundreds to tens of thousands of times compared with the current best solving algorithms.
The company specifically noted that this research focused on academic frontier schemes, historical competition candidate designs, and public-parameter challenges, and that the related findings do not affect the security of current mainstream cryptographic standards (such as the commonly used AES, RSA, SHA-256, etc.).
In accordance with the principles of responsible scientific research and security disclosure, MiniMax has synchronized the analysis results with the authors of the relevant schemes in advance, and has organized and verified the reasoning traces generated by MiniMax's internal model in the above experiments, making them public in the form of a technical report and code:
- https://github.com/MiniMax-AI/Research
Advertising statement: External redirect links contained in the article (including but not limited to hyperlinks, QR codes, passcodes, etc.) are used to convey more information and save selection time; the results are for reference only. All articles on IT Home containing external links include this statement.
Related Articles
- Day-0 support: Moore Threads completes adaptation for the MiniMax H3-Image image generation model
- MiniMax open-sources the H3-Image image generation model, generating a 4K image in 10 seconds
- MiniMax M3.1 Flash open-source model weights, with 428B total parameters
- DeepSeek and MiniMax have smoothly completed registration procedures and finished open-source and overseas authorized use
- Once up 7x, large-model company MiniMax's stock price falls below its IPO issue price
- MiniMax M3.1 Flash launches: 3x faster and 70% cheaper