Open challenge: pull the seed off an ESP32-S3 with Secure Boot + flash encryption in RELEASE (glitching allowed)

I’m the solo dev of Crypt-in, a file-encryption dongle on the ESP32-S3 (commercial, free beta for now). I’m posting here because the only attack I can’t test myself is the one this forum does best.

You flash a signed challenge image on your own S3. First boot burns Secure Boot v2 + flash encryption in RELEASE (XTS-AES-128), disables JTAG, sets anti-rollback, and generates a 32-byte seed on the chip. No command returns it, no recovery words. You send us the unit’s public key before you start; the flag is the seed that matches it. Glitching, power analysis and an opened package are all allowed on your own chip. To count, your method has to work on our reference chip too, so planting a seed before the seal doesn’t win.

Public state as of 11 Oct 2026: no known Secure Boot / FE bypass on the S3. AR2026-005 gets the general AES peripheral key, not the XTS key. AR2023-007 breaks C3/C6 and Espressif says it’s theoretically applicable to every XTS-AES chip, S3 included. The S3 has no hardware glitch detector.

Limits: no bounty (one-person project). You get credited by name, with a written statement of what your result means, and every claim and its verification gets published, failed ones too. Closes 31 March 2027.

Rules, reference keys, image hash: The extraction challenge | Crypt-in

Has anyone here pointed a Husky at an S3 yet? I’d like to know where the AR2023-007 approach falls apart on it, if it does.