Case Studies
How EmbedIDS is applied to concrete embedded and edge security problems, with the measurements to back it up.
Runtime security for Edge AI devices
EmbedIDS EdgeML detects rogue GPU workloads, AI model tampering and model exfiltration on a Jetson-class vision-inference device. It flags model tampering in 1 s and runs in about 9 MB of memory.
Read the Edge AI runtime security case study
Runtime intrusion detection on Zephyr RTOS
EmbedIDS runs inside Zephyr RTOS firmware as an optional module, with a bounded sample queue and one detector thread. In simulation it flags command abuse, request floods, tampering and telemetry loss within 120 ms, using 640 bytes of metric history.