TPM 2.0 Attestable Firmware: Recovering Trust After Firmware Flaws
summary
Shows how TPM 2.0 Library v183 adds new handles to CreatePrimary so the TPM mixes a bootloader-provided, version-unique secret into its DRBG and yields version-unique primary keys and a firmware-limited EK chained to the bootloader measurement (DICE-style). Implementation puts the bootloader in the reduced TCB, derives per-firmware secrets, and demo on Titan firmware seals data that fails to unseal after rollback, proving the recovery model works.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 0:47What is our plan when things go wrong
- 1:24How do you make sure that they landed
- 2:14How do we remotely verify
- 3:15How the TPM proves something
- 3:49Trusted Computing Base TCB
- 4:48Why is this bad
- 5:17Bad bugs
- 5:39Reducing TCB size
- 6:13Removing firmware from TCB
- 6:57In practice
- 8:39Create Primary
- 10:45New Handle Values
- 11:16Endorsement Key
- 12:08Public Key Harvesting
- 12:38Firmware Limited EAS
- 13:24Dice
- 14:31Demo
- 18:00Wrap Up
- 19:35Downgrade Protection
related talks
39mTPM Is No Silver Bullet: Pitfalls in Embedded Device Security
25mOverview and Recent Developments: TPM
33mfTPM: A Software-Only Implementation of a TPM Chip
51mApplications of TPM 2.0
42mTPMs and the Linux Kernel: unlocking a better path to hardware security
51mSecuring Embedded Linux Systems with TPM 2.0
42mRecent TPM Security Enhancements to the Linux Kernel
42mTPM based attestation - how can we use it for good?