Deep Learning for Blockchain Security: A Concise Survey, Taxonomy, and Open Problems

Blockchain is increasingly used in finance, healthcare, supply chains, and IoT, but it also faces growing security threats such as fraud, smart contract vulnerabilities, intrusions, and malware. Deep learning (DL) has emerged as a powerful tool to detect and mitigate these risks. This paper presents a structured survey of DL applications in blockchain security and proposes a taxonomy linking threat types, integration approaches (on-chain, off-chain, hybrid), DL models, data modalities, and deployment settings. We review recent advances across major security tasks and discuss key limitations and challenges. Finally, we highlight future research directions to improve robustness, efficiency, and practical deployment.