ELM: A Low-Latency and Scalable Memory Encryption Scheme

Memory encryption (ME) with authentication is becoming a key security feature of modern processors, as evident by the adoption ME Intel’s SGX. Recently actively studied from viewpoint system architecture. This paper studies symmetric-key cryptographic designs, primal focus on latency. A significant progress in such direction can be observed SGX Integrity Tree (SIT). Using variant AES-GCM, SIT achieves an excellent However, it has scalability issue. By carefully examining SIT, we develop new scheme dubbed ELM. We present AES-based instantiation ELM, and show that ELM significantly reduces latency for large memories, provable equivalent hardware-protected (on-chip) area. also preliminary hardware implementations to substantiate our advantages.