An Experimental Study of Merkle Tree-Based Security Mechanism for Secure SSD Storages

As modern applications rely on data-intensive workloads, adhering to intellectual value demands more memory than DRAM, which NVMe-based NAND storage fulfills. However, protecting these applications comes with significant additional metadata overhead and limitations. We evaluate secure memory-based SSD protection schemes, including the D-Shield-based design, and highlight the substantial performance degradation caused by frequent metadata requests required to maintain data security. Our experiments examine the performance of secure SSD configurations with and without metadata caching, revealing that caching helps reduce metadata requests by 25% and 60% in read and write-intensive workloads, respectively. Secure configurations still experience significant additional requests overhead of around 77% and a lack of on-drive integrity checks.