Learning Plaintext-Ciphertext Cryptographic Problems via ANF-based SAT Instance Representation
Authors: Xinhao Zheng, Yang Li, Cunxin Fan, Huaijin Wu, Xinhao Song, Junchi Yan
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, Crypto ANFNet achieves a 50x speedup over heuristic solvers and outperforms SOTA learning-based SAT solver Neuro SAT, with 96% vs. 91% accuracy on small-scale and 72% vs. 55% on large-scale datasets from real encryption algorithms. |
| Researcher Affiliation | Academia | Dept. of CSE & School of AI & Moe Key Lab of AI, Shanghai Jiao Tong University |
| Pseudocode | No | Section 4.4 is titled 'Key-solving Algorithm' and describes the algorithm's steps in paragraph form and through a pipeline diagram (Fig. 2), but it does not present it as a structured pseudocode block or a numbered algorithm listing. |
| Open Source Code | No | The source code will be made publicly available upon acceptance. |
| Open Datasets | No | For both training and test datasets, we evaluate our approach on two types of synthetic datasets. For the first synthetic dataset, we use a similar approach to Neuro SAT to generate instances of the MQ problem... For the second synthetic dataset, we utilize instances generated from the real encryption algorithms to construct the dataset... Depending on the differences in the encryption algorithms used, the datasets generated by this process are called Scipher-r-k-n and Speck-r-k-n. |
| Dataset Splits | No | For both training and test datasets, we evaluate our approach on two types of synthetic datasets. |
| Hardware Specification | Yes | We test all solvers on an AMD Ryzen Threadripper 3970X 32-Core Processor and an NVIDIA Ge Force RTX 3090 GPU. |
| Software Dependencies | No | The paper mentions external tools and solvers like WDSat [42], Bosphorus [43], Crypto Mini Sat [44], and Kissat [45], but does not provide specific version numbers for general software dependencies (e.g., programming languages, libraries like PyTorch or TensorFlow) used for implementing Crypto ANFNet. |
| Experiment Setup | Yes | In this experiment, we set λ = 0.1. |