Skip to content
Back to skills

Secure Boot Uefi Research

ASecurity

Boot-trust chain research: UEFI phases and variable services (PK/KEK/db/dbx), Secure Boot enforcement points, BCD policy persistence (testsigning/nointegritychecks), BitLocker key hierarchy and PCR binding, measured boot vs secure boot, bootkit precedents and their bypass classes, lab setup for OVMF/hardware research. Use when work touches boot trust, BitLocker, or pre-OS code.

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added October 17, 2026
securityrustgoshelldebuggingfrontendsecuritydocumentation

Security analysis

A100/100

Pro scans all 2 files and shows the line behind each finding

Scanned October 17, 2026

npx -y skills add NetVar1337/agent-skill-canon --skill secure-boot-uefi-research --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Secure Boot Uefi Research?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Secure Boot Uefi Research
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/netvar1337-secure-boot-uefi-research/badge)](https://www.skillsdirectory.com/skills/netvar1337-secure-boot-uefi-research)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: secure-boot-uefi-research
description: "Boot-trust chain research: UEFI phases and variable services (PK/KEK/db/dbx), Secure Boot enforcement points, BCD policy persistence (testsigning/nointegritychecks), BitLocker key hierarchy and PCR binding, measured boot vs secure boot, bootkit precedents and their bypass classes, lab setup for OVMF/hardware research. Use when work touches boot trust, BitLocker, or pre-OS code."
version: 1.0.0
license: GPL-3.0-or-later
metadata:
  package: unleash-skills
  author: NetVar1337/unleash
  category: stealth
  triggers:
    - "secure boot"
    - "uefi"
    - "bitlocker"
    - "bootkit"
    - "bcdedit"
    - "measured boot"
    - "tpm pcr"
---

# Secure Boot / UEFI research

The chain, in enforcement order: firmware (SEC→PEI→DXE→BDS) verifies the
boot loader (db signatures), bootmgr/winload verify kernel + ELAM + Critical
services, kernel CI takes over. Each link has distinct attack surface;
identify *which link* a question belongs to before touching anything.

## Inventory the box first (windows-internals platform truth)

```powershell
Confirm-SecureBootUEFI                       # SB state (needs admin, UEFI boot)
bcdedit /enum {current}                      # nointegritychecks/testsigning flags
manage-bde -status                            # BitLocker state, protector IDs
Get-Tpm; Get-TpmEndorsementKeyInfo           # TPM present/owned/attestation-ready
# Measured-boot evidence: parse C:\Windows\Logs\MeasuredBoot\*.log (TCG log) directly
```

UEFI variables live in `HKLM\SYSTEM\CurrentControlSet\Control\SecureBoot`
(state mirror) and the real store in NVRAM — query with
`GetUefiVariable`-style tools or the firmware shell; never assume the
registry mirror is authoritative.

## Variable/authority model

- `PK` (Platform Key) owns `KEK`; `KEK` updates `db` (allowed signers) /
`dbx` (revocation). `SetupMode` variable open at deployment = the classic
"deploy own keys" lab path (physical presence required).
- dbx updates arrive via Windows Update — for bootkit research, track dbx
  revocations as the defender's move history (each entry = a killed bypass).
- `*Default` variables matter: on some firmware, failure to read db falls
  back to vendor defaults — historically a bypass class (default-db
  trusting vendor-only chains).

## BCD policy layer (most-used in driver work)

```
bcdedit /set testsigning on            # allows test-signed kernel modules
bcdedit /set nointegritychecks on      # CI bypass for ELAM-era checks (may be ignored on recent builds)
bcdedit /set debug on                  # KDNET — pairs with windbg-ttd
```

These live in the BCD store (`\EFI\Microsoft\Boot\BCD` + registry hive
`HKLM\BCD00000000` view). Research notes: (a) flags are themselves measured
into PCR context on some configs (BitLocker sees tampering), (b) bootmgr
honors them before winload — a bootkit's cheapest trick is BCD surgery,
which is why defenders watch the store hash, (c) some flags are ignored when
Secure Boot is ON — always verify effective policy with an actual
unsigned-driver load test, not documentation.

## BitLocker key hierarchy

- FVEK (encrypts data) ← VMK ← protector(s): TPM (PCR-bound), PIN, USB,
  recovery password, AD escrow.
- PCR binding for TPM-only: PCR0..7 (firmware/config), PCR11 (bitlocker/
  access control). `testsigning` flips measured state → recovery prompt on
  next boot. Research flow: `manage-bde -protectors -get C:`, note PCRs,
  `suspend` before lab changes, `resume` after — avoid burning recovery
  prompts (they're logged and, on managed boxes, noticed).
- "Hardware encryption" on cheap SSDs (TCG Opal self-encrypting drives with
  default keys) historically degraded to trivially-decryptable states —
  check `manage-bde -status` "Encryption Method" before assuming FVEK-based
  crypto.

## Measured vs Secure Boot (don't conflate)

- Secure Boot = signature enforcement (an *active gate*).
- Measured Boot = PCR hash log (TPM) of everything booted — *evidence, not
  enforcement*. You can boot unsigned stuff and TPM happily records your sin;
  detection comes only from attestation (remote/PCR quotes — see
  `tpm-attestation-research`) or BitLocker's PCR-bound unsealing.
- TCG log (`\Windows\Logs\MeasuredBoot\*.log`) is parseable — as an attacker,
  read it to know exactly what you changed that got measured.

## Bootkit precedent classes (what public attacks did)

| Class | Mechanism | Lesson |
|---|---|---|
| MBR/VBR bootkits (pre-UEFI era) | stage0 in boot sector | dead on UEFI+GPT default installs |
| UEFI implant (LoJax-style) | DXE/runtime driver written to ESP/firmware | persistence below OS reimaging |
| Vulnerable-bootloader bypass (BlackLotus lineage) | abused a *signed-but-vulnerable* bootmgr/winload + dbx gaps to run unsigned policy | signing ≠ correctness; revocation lag is the window |
| BCD/policy surgery | flip nointegritychecks via physical/direct disk access | cheapest pre-OS tamper; measured |
| ESP implant | replace/chain bootloader on EFI System Partition | needs the partition writable — check BitLocker/ESP protections |

When researching a new one: diff the ESP contents (`mountvol S: /S`), hash
bootmgr/winload against known-good from the installed build, parse the TCG
log for unexpected measurements before the loader.

## Lab setups

- **QEMU + OVMF** (in WSL2/secondary box — QEMU not installed locally,
  install per need): enroll your own PK/KEK/db (`KeyTool` on the OVMF
  frontend), sign test binaries with created certs — full authority over
  every link; the only way to cheaply iterate on dbx/PKI logic.
- **Hardware**: Intel Boot Guard present boxes (OEM signed IBB) resist
  firmware replacement — check with `bootguard` status tools before planning
  firmware-level persistence experiments; boxes without Boot Guard remain
  open to SPI reflashing (CH341A-class programmer for extraction — pairs
  with `hardware-security`).
- Tooling gaps on this workstation: no UEFITool/chipsec — install when the
  task needs them and record in `tool-index.md`.

## OPSEC/tripwires while researching on live boxes

- Secure Boot state changes and BitLocker recovery events land in Event Log
  (Microsoft-Windows-BitLocker, Event 246/546 families) — clean your lab
  state *before* abandoning a box, or note the noise you made.
- Never flip secure-boot-critical vars on a box whose BitLocker recovery
  keys you don't hold.

## Pair with

`vbs-hvci-research` (the trust stack above boot), `tpm-attestation-research`
(PCR/attestation deep dive), `windbg-ttd` (boot debugging:
`bcdedit /bootdebug`), `firmware-pentest` (embedded/IoT firmware flows),
`hardware-security` (SPI/JTAG extraction), `windows-internals` (platform
truth first).

Files in this skill

  • SKILL.md6.6 KB
  • agents/openai.yaml43 B

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.