Optoelectronics Information Storage

Professor Yu Hua’s research group released new research in secure non-volatile memory research in HPCA 2021

Source:     The author:Hua Y    Delivery time:May 28, 2021    Browse:

The CCA A-level international conference HPCA 2021 recently published a paper entitled "Write-Friendly and Fast Recovery Solution for Secure Metadata in Non-Volatile Memory". The paper comes from Professor Yu Hua of Wuhan National Optoelectronics National Laboratory (WNLO), Huazhong University of Science and Technology.

It is important to ensure the security and integrity of data in non-volatile memory (NVM) . Existing work uses encryption schemes in NVM to construct an integrity tree and introduces many security metadata. However, after the system crashes and recovery, the security metadata in the cache will be lost, and the stale metadata in the NVM cannot protect the system. Existing secure metadata recovery schemes need to refresh all metadata, which will cause high write overhead and reduce system performance. Recovering security metadata requires a long recovery time and reduces system availability. In addition, attacks may occur during the recovery process and lead to incorrect recovery.

Figure1The overview of STAR design

In order to address these problems, the group of Professor Hua Yu of WNLO proposed a low write overhead, fast and correct recovery scheme, called STAR, as shown in Figure 1. STAR proposes counter-MAC synergization to store modifications to metadata without writing additional memory. In order to recover quickly, STAR uses bitmap lines to record the location of the outdated metadata, and only needs to restore the outdated metadata, not all metadata. STAR proposed a cache tree to verify the recovery process. As shown in Figure 2, compared with the state-of-the-art schemes, STAR has reduced additional memory writes by 92%.

Figure2The write traffic of STAR in different workloads

This work was supported in part by National Key Research and Development Program of China under Grant 2016YFB1000202, and National Natural Science Foundation of China (NSFC) under Grant No. 61772212. Professor Yu Hua from WNLO is the corresponding author and the Ph.D student Jianming Huang is the first author.