Confidential In-Memory Computing Architectures for Scalable Real-Time Cloud Security Analytics
Year: 2023
Type: article
Abstract: Modern cloud environments generate massive volumes of high-velocity telemetry data (e.g., network flows, system logs, API call traces) requiring real-time threat analysis. Processing this data securely and at low latency introduces a fundamental trade-off: traditional cryptographic boundaries severely impair execution throughput, while plain-text processing exposes sensitive operational data to insider threats and hypervisor compromise. This paper proposes HyperSec-Engine, a distributed, high-performance in-memory analytics architecture designed to process encrypted telemetry streams at line rate. By pairing Hardware-Assisted Execution Environments (TEEs) with Remote Direct Memory Access (RDMA) and zero-copy data structures, HyperSec-Engine achieves near-native analytical throughput while maintaining zero-trust confidential processing guarantees. Empirical evaluations demonstrate an analytical throughput of 19.2 Million events/sec per node and 92.15% linear scaling efficiency across a 16-node cluster, outperforming baseline TEE implementations by 5.3x in throughput and 7.0x in p99 tail latency.
DOI: 10.69889/1hmk4132
Author: Dr. Narendra Chaudhari
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W7204654178