CyberPeople

Flash-Chip Attack Bypasses TPM Lockout: Windows Hello and BitLocker PINs Cracked in Hours

TPM Lockout Bypass Cracks Windows Hello and BitLocker PINs

TU Berlin researchers bypassed the TPM lockout and cracked Windows Hello and BitLocker PINs in hours instead of years. Here's how the attack works and how to defend.

What the TPM is and why a short PIN is safe at all

A short numeric PIN looks like a weak password on the surface. Four digits give only 10,000 combinations, which could be brute-forced in minutes. But Windows Hello makes such a PIN effectively secure thanks to the hardware protection of the TPM (Trusted Platform Module). It is the TPM that implements anti-hammering: after a number of failed attempts it stops responding, so an attacker cannot simply grind through PINs one after another.

The scale here is enormous. Windows Hello is the primary authentication mechanism on more than a billion Windows 11 devices, and it allows PINs as short as four digits. That low entropy is only acceptable because the TPM strictly limits the number of attempts. By Microsoft's defaults, the TPM locks after 32 failed tries and forgets one failure every 10 minutes. This is the mechanism that turns a weak PIN into a solid barrier.

Now that guarantee is in question. Researchers at Technische Universität Berlin published "Hello Again: Exploiting Replay Protected Monotonic Counters to Brute-Force Windows Hello PINs", in which they showed that physical access to a laptop lets an attacker "rewind" the failed-attempt counter and brute-force a Windows Hello or BitLocker PIN in hours — rather than the years the design intended.

Firmware TPM and the flash chip: where the counter lives

To understand the attack, it helps to see how the TPM is built in modern laptops. On most x86 platforms the TPM is implemented not as a separate chip but as a firmware TPM (fTPM) — part of the processor's firmware. This is cheap and convenient for manufacturers, but it has one fundamental consequence: unlike a discrete TPM chip, which has its own secure storage, an fTPM has none.

So its state between reboots — including the failed-attempt counter and the lockout state — is stored in encrypted form on the ordinary flash chip on the motherboard, connected over the SPI bus. And that flash chip turned out to be the weak point.

The attack idea is simple and elegant. Take a snapshot of the fTPM's saved state (with the current attempt counter), try a few PINs, then write the old copy back — and the counter "rolls back". The computer will believe no failed attempts happened and will grant a fresh set of tries. To stop this, newer platforms use RPMC (Replay Protected Monotonic Counters) — counters inside the flash chip that can only increase and cannot be rolled back. This is a JEDEC-standardized mechanism, and it was supposed to stop rollback attacks.

First practical analysis of RPMC — and two flaws at once

First practical analysis of RPMC — and two flaws at once

The Berlin team performed the first practical security analysis of RPMC in real, shipping x86 systems. To do this they built low-cost interposer hardware — a board placed physically between the processor and the flash chip. The board intercepts SPI traffic and can manipulate it. The result: two fundamental flaws, each of which breaks RPMC's security model on its own.

The first flaw is in Winbond flash chips, the leading NOR flash vendor with over 25% market share. Under a brief drop in power supply — a voltage glitch — a Winbond chip leaks the secret HMAC key that signs its counter values. With that key, which took about a second to extract at their best settings, the board can forge any counter value. Rollback protection vanishes entirely.

The second flaw is in AMD's firmware. Its NV controller (the component responsible for non-volatile memory) never verifies that the RPMC counter actually increased. So the interposer can simply swallow the counter-update command while reporting success — and the platform accepts the old state without suspicion. On top of that, AMD platforms without RPMC flash appear to lack NV rollback protection altogether: on two AMD desktops, simply reflashing an old firmware image reset the fTPM's lockout counters.

Both flaws allow rolling back the fTPM's state — whether AMD, Intel, or Microsoft Pluton — including its lockout counters. And that is a direct path to brute-forcing the PIN.

How long PIN cracking now takes

How long PIN cracking now takes

The numbers the research team reports are striking. For a four-digit Windows Hello PIN, the worst-case brute-force time drops from 69 days (with lockout working as documented) to 40 hours. For a default six-digit BitLocker PIN, it drops from 19 years to 20 days. In other words, protection that used to buy years now collapses to hours or days.

The team confirmed this end-to-end on real, encrypted hardware. For the Windows Hello test they used a locked HP OmniBook 3 with an AMD Ryzen AI 5 330 processor and a BitLocker-encrypted drive. Their setup started typing PINs from 1000 upward and reached the correct one — 1235 — after 236 tries in 57 minutes. Notably, Windows forced a restart after every 5 wrong attempts on that laptop, but no lockout delay was applied. On an Intel laptop, a six-digit BitLocker PIN was recovered in under seven minutes — also after 236 attempts.

The consequences go well beyond logging in. A compromised PIN unlocks device-bound keys and Hello-based enterprise SSO — meaning an attacker who cracks the PIN can reach a company's internal systems without ever breaking the user's password.

AMD and vendor response

AMD confirmed the flaw in its firmware and assigned it CVE-2026-43613 with a CVSS score of 5.1. Security bulletin AMD-SB-7079 was published on November 18, 2026. The vendor notes that attacks requiring physical access generally fall outside its product security threat model, but at the same time acknowledges that "RPMC is intended to stop this type of interposer attack". The researchers first reported the issue to AMD back in February. Fixes are planned in stages — from November 2026 through February 2027 — and these are plans, not updates you can install today.

The Winbond flash situation is trickier. The voltage-glitching vulnerability is a separate issue, and no confirmed fix has been announced at the time of publication. For users this means millions of laptops remain exploitable right now.

How to defend

How to defend

The simplest and most effective recommendation from the researchers themselves is to make the PIN longer. Add at least two digits to your Windows Hello and BitLocker PINs: that pushes brute-force difficulty back up by orders of magnitude even when the lockout is bypassed. Better still, use long alphanumeric PINs or full passwords instead of short numeric sequences.

Organizations should revisit their endpoint security policies. Do not rely on a PIN alone to protect critical data, especially on devices that can be lost, stolen, or accessed by an insider. For the most sensitive systems, consider discrete hardware TPM chips and long BitLocker passwords that resist brute force. Physical access is the main precondition for the attack, so access control over hardware — in the office, during repair, or across the supply chain — matters just as much as configuration.

This finding is another reminder that security starts not only with software but with understanding the hardware mechanisms underneath. If you want to systematize your knowledge of hardware security, take a look at cybersecurity courses and training programs — they help clarify how mechanisms like the TPM actually work and why they sometimes fail.

Conclusion

The Berlin team's work is not just another entry in a long list of vulnerabilities. It strikes at the fundamental assumption on which the security of hundreds of millions of devices rests: that a short PIN is safe because the TPM protects it. It turns out that protection can be bypassed with an inexpensive board and physical access, and short PINs can be cracked in hours instead of years. AMD's fixes are stretched out to February 2027, and the Winbond issue has no confirmed fix at all. So the most reliable defense today is a long, complex PIN or password that stays strong even when the hardware protection fails.

Stay ahead of threats

Weekly cybersecurity intelligence in your inbox. No spam.

CyberPeople contributor