Privacy-Aware Emerging Computing FIGURE 5. Performance comparison of the stacked neural network with the state-of-the-art models for the power-system industrial dataset. FIGURE 7. Overall architecture of the JARA IDS. and perform traffic discrimination. JARA has been developedto bedeployedon a single host/sensor and perform threat analysis for a specific network interface or prelogged packet captures. With the learning procedure of the neural-network stack discussed in this article, the proposed IDS JARA12 effectively detected intrusions (see Figure 8) with appropriate IP-evidences and behavior profiles without runtime FP during initial prototype testing when deployed on a Linux virtual machine and validated against the persistent reboot-survival IIoT malware hide andseek.13 FIGURE 6. Performance comparison of the stacked neural network with the state-of-the-art models for the Edge-IIoTset dataset. tive phases, where the first/training-phase groundtruth of the network traffic is captured to label the packet sequences. In the second/validation phase, the configuration parameter of JARA will be set to " Test " mode for analyzing the real-time network Architectural Overview The overall architecture of JARA is illustrated in Figure 7. Each network traffic imputed is first segmented into IP-address-based packet chunks. These chunks are then profiled based on time window, packet sequencing,and protocol followed. These profiles capture utmost traffic behavior within a simple data structure FIGURE 8. Screenshot of JARA IDS at runtime. 36 IEEE Consumer Electronics Magazine