arXiv cs.CR·9d agoCESBench: Benchmarking Large Language Models on Cryptographic Engineering Security for IoT Devices#cesbench#llm#benchmarkAI research