FPGA-Accelerated GNN Pipelines for Real-Time Identity Threat Scoring in Zero-Trust Cloud Systems
Year: 2025
Type: article
Abstract: Zero-trust cloud systems demand real-time, continuous identity threat assessment of all access requests, with the restrictive latencies and determinism limits that cannot be satisfied to scale by software-only security pipelines. Identity threat scoring is based on high-speed computation of multifaceted relationship graphs among users, devices, roles, and services, which form bottlenecks in performance when the occurrence rate of event-consumption patterns is large and times of fluctuating access levels. In this paper, a system architecture of real-time identity threat scoring using FPGA is given, embedded as a security control path in a zero-trust cloud. The design has the additional advantage with a deterministic execution of an offload mechanism to a streaming hardware pipeline that offloads the graph traversal operations and scoring operations, both of which are latency critical, to the streaming hardware pipeline to support high throughput. Hardware-software co-designs represent a separation of control-plane administration and data-plane implementation, guaranteeing safe and solid exploitation. The lack of experimental results on a cloud-based testbed indicates that control charts on the reduction of end-to-end decision latency, enhancement of improved throughput in the event of a burst workload are achieved, and the predictability of performance is credited to FPGA-based acceleration due to its practicability in zero-trust identity enforcement such as software-only implementation.
Author: O.J.M. Smith
Institution: Universidade Federal de Pernambuco
Language: English
Subfield: Hardware and Architecture
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7219942006
