Anyone Can Hack IoT Even Easier Now - A Beginner's Guide to AI Augmented IoT Hacking
Where
IoT Village
DEF CON 34
Duration
40 min
Shows how a beginner can use AI agents and a cheap bench rig to find and exploit bugs in consumer IoT firmware, using a router as the example.
Summary
The speaker revisits his earlier DEF CON talk on IoT hacking and argues the barrier has dropped further now that AI agents can handle much of the groundwork. He lays out a beginner workflow: a kit of about twenty dollars (a multimeter, a USB-to-serial adapter and a flash programmer), then a cheap router or camera running embedded Linux as the target. Firmware can often be downloaded from vendor support pages, or read off the board when it is not.
He then separates the tasks AI does well from those that still need a person. Models are strong at reading scripts, logs and decompiled code, but they misjudge real-world severity and tend to chase privilege escalation on single-user root systems. Soldering and probing remain manual. He presents Wairz, an open-source MCP server he built that unpacks firmware in Docker and gives an agent reverse engineering, taint analysis, emulation and fuzzing tools. He also describes a fleet of Raspberry Pis that lets agents reach serial consoles, power relays, Wi-Fi access points, BLE, a rooted Android phone and SPI or JTAG interfaces.
As a test, an agent picked a cheap router, recovered its UART password, traced a custom web server to an unsanitized system call and a stack overflow, and built a working exploit that the speaker verified by hand. That yielded three CVEs. He closes with cautions about hallucinations, data handling and authorization, and points to a free UART course and a hands-on workshop.
Materials
Related talks
IoT Village·43 min
Barak Sternberg
Explores vulnerability discovery techniques and exploitation workflows targeting connected smart devices.
BioHacking Village·32 min
Marie Mo · Cheyenne · Christian
Follows a pacemaker patient and two students as they reverse engineer a clinic programmer and an implant's firmware, exposing its weaknesses.
IoT Village·42 min
Larry Pesce
Shows how an AI agent driven by an IoT pentest playbook found and chained flaws in a webcam, and what humans must still validate.


