LLMs: Your Wireless Hacking Friend

Alissa Gilbert argues that general-purpose chatbots can shorten parts of wireless penetration testing. She starts with the basic flow of recon, enumeration and attack, then outlines what language models handle well: reading logs, interpreting packet captures and helping with reconnaissance. She is clear about their limits. They cannot run attacks or touch hardware, and prompts work better when they state a lab or testing context, stay narrow, and include device details or documentation.

The middle of the talk walks through example prompts. They cover a Wi-Fi reverse shell draft, retrieving CVE details for a router model, encrypting payloads, finding a four-way handshake in a capture, and spotting hidden SSIDs in packet data. Generated output is then fed into aircrack-ng and other tools. She also discusses wordlist suggestions for WPA-PSK cracking and a WPA3 Dragonblood prompt. Her tips focus on referencing timeframes or packet numbers and splitting large captures.

She closes by noting that her attempts with a self-hosted model fell short, and that she prefers Claude for scripting. The takeaway is that these tools speed up familiar steps, but an experienced operator still has to check results, reprompt when answers drift, and verify output before relying on it.