A Blockchain-Enabled Deep Learning Framework for Secure Omics Data Sharing and Attack Detection

Abstract

Securing omics datasets against tampering and misuse is essential for reproducible research and privacy. We present a defense-in-depth framework that combines Ethereum smart contracts for tamper evidence, provenance, and fine-grained access with a Long Short-Term Memory (LSTM) intrusion detection system that models event sequences to flag abnormal behavior. Raw omics files remain off-chain; their SHA-256 digests and permissions are recorded on-chain. Authorized consumers obtain contract-mediated tokens to fetch encrypted data, recompute hashes, and verify integrity. The intrusion detection system (IDS) ingests blockchain transactions and storage access logs in sliding windows to detect bursts, probing, and insider over-access. We implement the system on a permissioned Ethereum network and evaluate it with a simulated case study using public gene-expression files and scripted attacks. All post-registration data modifications were detected at verification time (100% integrity detection). Behavioral attacks were identified with 95% precision and 90% recall, reducing false alarms to 1% of windows and outperforming a rules-only baseline (80% precision, 75% recall). Transaction latency and resource costs remained modest. These results demonstrate a practical path to trustworthy omics sharing that unites cryptographic immutability with monitoring. Our design supports consent enforcement and lays groundwork for extensions such as Merkle-root batching, key rotation, and federated or transformer-based detectors.