Everything I own, owned

A researcher used the AI model Claude Opus 5 to reverse‑engineer the firmware of five USB and Wi‑Fi peripherals, uncovering serious security flaws in each device. By extracting firmware and update tools from the manufacturers and feeding them to Claude, the researcher obtained a plaintext command shell inside a microphone, disabled the activity LED on an Insta360 Link webcam while it continued recording, and discovered that a key light allowed unauthenticated memory writes over the network. The process involved prompting Claude with detailed analysis goals—such as documenting update formats, implementing custom utilities, and enumerating all protocol surfaces—and resulted in roughly 13 hours of AI “churn” and 98 user prompts across two weeks.

The most consequential finding was the ease with which the devices could be compromised. The Insta360 webcam runs an Ambarella‑supplied ThreadX RTOS and exposes a USB vendor‑class command channel that permits arbitrary file reads, writes, and reboots, enabling a full firmware flash without user interaction. Its firmware only checks an appended MD5 hash, allowing the researcher to patch the LED control table, recalculate the hash, and flash the modified firmware so the recording indicator never lights. Similarly, the ASUS ROG Swift PG42UQ monitor’s firmware uses a simple A/B slot system with a basic checksum, offering no protection against arbitrary writes over its I2C‑over‑USB update path, which the researcher could exploit to suppress unwanted pixel‑cleaning overlays.

All five devices now have publicly available GitHub repositories containing the generated documentation, scripts, and tools used in the analysis, validated on real hardware. The work demonstrates that peripheral firmware often lacks robust authentication or secure boot mechanisms, exposing users to covert surveillance and unauthorized control. As the researcher notes, these findings highlight a broader attack surface in everyday accessories, suggesting that manufacturers need to adopt stronger cryptographic verification and tamper‑resistance to protect against similar AI‑assisted exploitation.

⚡ Effects Interpreter