faulTPM: Exposing AMD fTPMs' Deepest Secrets
Listen
Radio episode about this paper
Transcript
Introduction to the show: ident: Security Radio. Generated commentary on the latest security and cryptography papers.
Nadia: I'm Nadia.
Elias: And I'm Elias, here with Priya.
Nadia: We're looking at the paper: "faulTPM: Exposing AMD fTPMs' Deepest Secrets".
Elias: Here's the gist of the paper. Trusted Platform Modules (TPMs) are vulnerable to physical attacks against their execution environment, which can lead to a full TPM state compromise This vulnerability allows an attacker to extract any cryptographic material stored or sealed by the fTPM regardless of authentication mechanisms This attack can be used to successfully attack Microsoft BitLocker’s TPM-only key protector Attack Mechanism and Impact The paper demonstrates that AMD’s fTPMs are vulnerable to physical attacks against their execution environment, the AMD Secure Processor (AMD-SP) The attack utilizes the AMD-SP’s vulnerability to voltage fault injection attacks to extract a chip-unique secret from the targeted CPU This secret is subsequently used to derive the storage and integrity keys protecting the fTPM’s non-volatile (NV) data stored on the Basic Input/Output System (BIOS) flash chip In contrast to previous dTPM sniffing attacks, this approach exposes the complete internal TPM state of the fTPM Furthermore, once we have extracted the chip-unique secret through voltage fault injection, we only need to re-read the BIOS flash to defeat any TPM-based security measures Key Derivation and Extraction The authors reverse-engineered the NV storage format of AMD’s fTPM and the derivation of the chip-unique keys protecting its confidentiality and integrity The non-volatile (NV) state of AMD’s fTPM is stored on the motherboard’s BIOS flash chip The storage and integrity keys are derived from a 128 bit secret unique to each CPU This derivation process involves a sequence of steps including AES128 decryption, KDF3 application, and SHA256 HMAC calculations The chip-unique secret is contained in a read-protected area of the protected slot Case Study on FDE Security The paper evaluates the severity of its fTPM attack by conducting a case study on BitLocker’s TPM-only and TPM and PIN protectors The results show that BitLocker’s TPM-only protector is ineffective when the TPM’s internal state can be extracted In contrast, BitLocker’s TPM and PIN protector retains the protection that a PIN-only strategy would offer The authors find that a TPM and PIN protector can be implemented such that the security of a PIN-only strategy remains once the fTPM state is compromised Mitigation Strategies Stopping arbitrary code execution attacks on the AMD-SP is identified as the only way to mitigate our attack effectively For Zen 2/3 systems, mitigations involve changes in the hardware of the SoC or the ROM boot loader of the AMD-SP Intel’s introduction of digital detection circuitry into their Converged Security and Manageability Engine (CSME) is mentioned as a preemptive measure The paper concludes that users who fear a physical attacker with reasonable resources should opt for a TPM and PIN configuration When BitLocker identifies that the underlying TPM is an fTPM, users should be urged to turn their PIN into a passphrase Awareness of the required security levels and the TPM variant in use is essential Data Availability The authors publish all code necessary to mount the attack under
27: The repository further includes several intermediate results, e.g., flash memory dumps, to retrace the attack process without possessing the target boards and required hardware tools The paper's arXiv ID is 2304.14717.
Nadia: References
2: Advanced Micro Devices.
Elias: AMD Firmware TPM Updates
3: Advanced Micro Devices.
Nadia: AMD Memory Guard - Whitepaper
4: Advanced Micro Devices.
Paper summary: Nadia: We’re talking about this paper now, "faulTPM: Exposing AMD fTPMs' Deepest Secrets." The central thesis is pretty direct: Trusted Platform Modules are vulnerable to physical attacks against their execution environment, which can lead to a full TPM state compromise <ref:2304.14717#pg1>.
Elias: They claim this vulnerability lets an attacker extract any cryptographic material stored or sealed by the fTPM, no matter what authentication mechanisms you have in place <ref:2304.14717#pg2>. This is significant because it bypasses things like PCR validation or passphrases with anti-hammering protection.
Priya: So, to put it simply for the listener, this paper argues that attackers can get a complete picture of everything the TPM has sealed inside, not just what's allowed to be used.
Nadia: Right. They verify this by compromising the AMD Secure Processor, which is the Trusted Execution Environment for these fTPMs <ref:2304.14717#pg2>. By attacking this processor, they extract a chip-unique secret from the CPU <ref:2304.14717#pg2>.
Elias: That secret then lets them derive the storage and integrity keys protecting the non-volatile data on the BIOS flash chip <ref:2304.14717#pg2>. They show that this is a new class of attack against fTPMs, different from previous dTPM sniffing attacks <ref:2304.14717#pg1>.
Priya: What does this mean for the security landscape? If you use an fTPM, does that automatically put you at risk of this type of compromise?
Nadia: It means that relying on an fTPM alone isn't enough if you’re worried about physical access <ref:2304.14717#pg3>. This attack shows a way to defeat TPM-based security measures when the internal state is compromised, which we see in case studies with BitLocker <ref:2304.14717#pg3>.
Elias: The paper highlights that even with a TPM and PIN protector, if the fTPM state is compromised, it retains the protection that a PIN-only strategy would offer <ref:2304.14717#pg3>. That’s an important distinction for how you think about layered security.
Priya: So, to recap: they found a way to compromise the entire internal state of an fTPM by targeting its execution environment, which lets them extract secrets and potentially defeat standard protection methods <ref:2304.14717#pg1>.
Conclusion: Nadia: So, looking at the whole paper, "faulTPM: Exposing AMD fTPMs' Deepest Secrets," the authors are basically showing that physical access can be a serious threat to these components <ref:2304.14717#pg1>.
Elias: The authors are Hans Niklas Jacob, Christian Werling, Robert Buhren, and Jean-Pierre Seifert from Technische Universitat Berlin <ref:2304.14717#pg1>. Their work focuses on exploiting the execution environment of AMD fTPMs <ref:2304.14717#pg2>.
Priya: The implication for us is that we need to be aware of the specific TPM variant in use and what security level you actually need for your data at rest <ref:2304.14717#pg3>. This isn't just abstract theory; it’s about practical choices.
Nadia: Exactly. The paper suggests that if you fear a physical attacker with reasonable resources, the best bet is to use a TPM and PIN configuration <ref:2304.14717#pg3>. If you are using an fTPM and BitLocker detects it, you should turn your PIN into something more like a passphrase <ref:2304.14717#pg3>.
Elias: Ultimately, the paper demonstrates that if an attacker can compromise the internal state of the fTPM through voltage fault injection, any TPM-based security is potentially defeated <ref:2304.14717#pg3>. It’s a strong message about physical security versus software protection.
Priya: It really shows that securing data at rest requires considering both the software layer and the physical layer of defense <ref:2304.14717#pg3>. That balance is something we should all be thinking about when choosing our security implementations.
Technische Universitat Berlin · Fraunhofer SIT
cs.CR
Submitted: 2023-04-28
Updated: 2026-11-05
Comments: The first two authors contributed equally. We publish all code necessary to mount the attack under https://github.com/PSPReverse/ftpm_attack. The repository further includes several intermediate results, e.g., flash memory dumps, to retrace the attack process without possessing the target boards and required hardware tools
DOI: 10.1109/EuroSP57164.2023.00069
Code: https://github.com/Aorimn/dislocker
License: http://creativecommons.org/licenses/by-sa/4.0/
Importance score: 99/100
The gist: Trusted Platform Modules (TPMs) are vulnerable to physical attacks against their execution environment…
Key concepts
- fTPM
- Firmware TPM is a type of Trusted Platform Module integrated into the system's firmware. Unlike discrete TPMs, fTPMs do not communicate over an exposed bus to the CPU, making them immune to certain sniffing attacks. However, this paper shows that attacking their internal execution environment can compromise the entire state.
- AMD-SP
- The AMD Secure Processor is the Trusted Execution Environment (TEE) for AMD's fTPMs. The attack targets vulnerabilities within this processor using voltage fault injection to extract a secret key unique to each CPU. This secret is crucial because it protects all the non-volatile data stored on the system.
- Voltage Fault Injection
- This is a physical attack technique where precise, temporary fluctuations in the voltage supplied to a chip are induced. The goal is to cause transient errors during critical operations within the AMD-SP. By inducing these faults, attackers can force the processor to leak sensitive information, such as the chip-unique secret.
- TPM State Compromise
- This means an attacker gains complete access to all data and secrets stored or sealed within the TPM. Because this attack extracts keys used for storage and integrity, it defeats security measures like Platform Configuration Register (PCR) validation, allowing extraction of cryptographic material regardless of how the system is authenticated.
Terminology
Summary
Trusted Platform Modules (TPMs) are vulnerable to physical attacks against their execution environment, which can lead to a full TPM state compromise This vulnerability allows an attacker to extract any cryptographic material stored or sealed by the fTPM regardless of authentication mechanisms This attack can be used to successfully attack Microsoft BitLocker’s TPM-only key protector
Attack Mechanism and Impact
The paper demonstrates that AMD’s fTPMs are vulnerable to physical attacks against their execution environment, the AMD Secure Processor (AMD-SP) The attack utilizes the AMD-SP’s vulnerability to voltage fault injection attacks to extract a chip-unique secret from the targeted CPU This secret is subsequently used to derive the storage and integrity keys protecting the fTPM’s non-volatile (NV) data stored on the Basic Input/Output System (BIOS) flash chip In contrast to previous dTPM sniffing attacks, this approach exposes the complete internal TPM state of the fTPM Furthermore, once we have extracted the chip-unique secret through voltage fault injection, we only need to re-read the BIOS flash to defeat any TPM-based security measures
Key Derivation and Extraction
The authors reverse-engineered the NV storage format of AMD’s fTPM and the derivation of the chip-unique keys protecting its confidentiality and integrity The non-volatile (NV) state of AMD’s fTPM is stored on the motherboard’s BIOS flash chip The storage and integrity keys are derived from a 128 bit secret unique to each CPU This derivation process involves a sequence of steps including AES128 decryption, KDF3 application, and SHA256 HMAC calculations The chip-unique secret is contained in a read-protected area of the protected slot
Case Study on FDE Security
The paper evaluates the severity of its fTPM attack by conducting a case study on BitLocker’s TPM-only and TPM and PIN protectors The results show that BitLocker’s TPM-only protector is ineffective when the TPM’s internal state can be extracted In contrast, BitLocker’s TPM and PIN protector retains the protection that a PIN-only strategy would offer The authors find that a TPM and PIN protector can be implemented such that the security of a PIN-only strategy remains once the fTPM state is compromised
Mitigation Strategies
Stopping arbitrary code execution attacks on the AMD-SP is identified as the only way to mitigate our attack effectively For Zen 2/3 systems, mitigations involve changes in the hardware of the SoC or the ROM boot loader of the AMD-SP Intel’s introduction of digital detection circuitry into their Converged Security and Manageability Engine (CSME) is mentioned as a preemptive measure
Conclusion on User Choice
The paper concludes that users who fear a physical attacker with reasonable resources should opt for a TPM and PIN configuration When BitLocker identifies that the underlying TPM is an fTPM, users should be urged to turn their PIN into a passphrase Awareness of the required security levels and the TPM variant in use is essential
Data Availability
The authors publish all code necessary to mount the attack under [27] The repository further includes several intermediate results, e.g., flash memory dumps, to retrace the attack process without possessing the target boards and required hardware tools The paper's arXiv ID is 2304.14717.
References
[2] Advanced Micro Devices. AMD Firmware TPM Updates
[3] Advanced Micro Devices. AMD Memory Guard - Whitepaper
[4] Advanced Micro Devices. Whitepaper: AMD Ryzen Pro 5000 Series Mobile Processor Security Features
[5] E. Agostini and M. Bernaschi. BitCracker: BitLocker meets GPUs
[6] Paul Alcorn. AMD Sets All-Time CPU Market Share Record as Intel Gains in Desktop and Notebook PCs
[7] Denis Andzakovic. Extracting BitLocker keys from a TPM
[8] Aorimn. Dislocker
[9] Will Arthur, David Challener, and Kenneth Goldman. A Practical Guide to TPM 2. p. 1 However, the authors
neglected to analyze the impact of this Root of Trust
(RoT) compromise on other critical security functionality
provided by the AMD-SP. In light of these attacks, we ask: What are the practical
consequences of physical attacks against AMD’s fTPMs?Where the paper says it
[10] Benjamin Bock. Firewire-based Physical Security Attacks on Windows 7, EFS and BitLocker
[11] Milan Broz. LUKS2 On-Disk Format Specification
[12] Robert Buhren and Alexander Eichner. All you ever wanted to know about the AMD Platform Security Processor and were afraid to emulate
[13] Robert Buhren, Alexander Eichner, and Christian Werling. Uncover, Understand, Own - Regaining Control Over Your AMD CPU
[14] Robert Buhren, Hans Niklas Jacob, Thilo Krachenfels, and JeanPierre Seifert. One Glitch to Rule Them All: Fault Injection Attacks Against AMD’s Secure Encrypted Virtualization
[15] Robert Buhren and Niklas Jacob. Glitching the AMD Secure Processor
[16] Robert Buhren, Christian Werling, and Jean-Pierre Seifert. Insecure Until Proven Updated: Analyzing AMD SEV’s Remote Attestation
[17] William E. Burr et al. Electronic Authentication Guideline
[18] Lily Chen. Recommendation for key derivation using pseudorandom functions (revised)
[19] Cfir Cohen. Full Disclosure: AMD-PSP: fTPM Remote Code Execution via crafted EK certificate
[20] Thomas Dewaele and Julien Oberson. TPM sniffing – Sec Team Blog
[21] Grigori Goronzy. Cryptenroll: Add support for TPM2 pin
[22] Seunghun Han et al. A Bad Dream: Subverting Trusted Platform Module While You Are Sleeping
[23] Chris Hoffman. How to Enable a Pre-Boot BitLocker PIN on Windows
[24] Intel Corporation. Whitepaper: Strengthening Security with Intel Platform Trust Technology
[25] Intel Corporation. Whitepaper: Intel® Total Memory Encryption
[26] Intel Corporation. Code Sample: Protecting secret data and keys using Intel® Platform Trust Technology
[27] Hans Niklas Jacob and Christian Werling. PSPReverse/ftpm attack on Github
[28] Roger Lai. AMD Security and Server innovation
[29] Microsoft. BitLocker - Windows security
[30] Microsoft. BitLocker Countermeasures (Windows 10) - Windows security
[31] Microsoft. How Windows uses the TPM - Windows security
[32] Microsoft. Kernel DMA Protection (Windows) - Windows security
[33] Microsoft. Prepare your organization for BitLocker Planning and policies
[34] Microsoft. Trusted Platform Module (TPM) fundamentals (Windows)
[35] Microsoft. Windows 11 Specs and System Requirements — Microsoft
[36] Microsoft. Manage-bde protectors
[37] Daniel Moghimi et al. TPM-FAIL: TPM meets Timing and Lattice Attacks
[38] Daniel Nemiroff and Carlos Tokunaga. Tunable Replica Circuit for Fault-Injection Detection
[39] Daniel Nemiroff and Carlos Tokunaga. Whitepaper: Fault-Injection Countermeasures, Deployed at Scale
[40] OASIS PKCS 11 Technical Committee. PKCS 11 Cryptographic Token Interface Base Specification Version 2. p. 2 A significant application of TPMs
is hardware-aided key management. The tpm2-pkcs11
project [46], for example, exposes the TPM’s functionality
as a PKCS #11 interface, which is a standardized API to
access cryptographic services like smartcards [40].Where the paper says it
[48] Trusted Computing Group. TPM 2.0 A Brief Introduction
[50] Johannes Winter and Kurt Dietrich. A hijacker’s guide to communication interfaces of the trusted platform module
[51] Philip Yip. BitLocker Installation Guide
[52] Zeljka Zorz. Intel releases updates to plug TPM-FAIL flaws
--- Page 1 ---
faulTPM: Exposing AMD fTPMs’ Deepest Secrets
Abstract—Trusted Platform Modules (TPMs) constitute an integral building block of modern security features. Moreover, as Windows 11 made a TPM 2. p. 0 de
Abstract—Trusted Platform Modules (TPMs) constitute an
integral building block of modern security features. Moreover, as Windows 11 made a TPM 2.Where the paper says it
Pruning redundant reasoning tokens
The paper analyzes a new class of attacks against fTPMs: Attacking their Trusted Execution Environment (TEE) can lead to a full TPM state compromise This attack exposes the complete internal TPM state of the fTPM It allows us to extract any cryptographic material stored or sealed by the fTPM regardless of authentication mechanisms such as Platform Configuration Register (PCR) validation or passphrases with anti-hammering protection
Hardware Attack Details
The attack leverages the AMD-SP’s vulnerability to voltage fault injection attacks to extract a chip-unique secret from the targeted CPU This secret is subsequently used to derive the storage and integrity keys protecting the fTPM’s non-volatile (NV) data stored on the Basic Input/Output System (BIOS) flash chip The attack requires access to the motherboard of the target system, particularly its SPI bus and voltage regulators Hardware setup requires an assortment of readily available hardware such as an SPI Programmer and a Logic Analyzer
Non-Volatile Storage Analysis
The non-volatile (NV) state of AMD’s fTPM is stored on the motherboard’s BIOS flash chip The NV DATA file is divided into two 64 KiB sections featuring append-only data structures Each entry consists of a variably sized body encrypted using AES128 in counter mode The storage and integrity keys are derived from a 128 bit secret unique to each CPU
Case Study on BitLocker
The authors conducted a case study on the implementation of BitLocker’s TPM-only and TPM and PIN protectors They found that BitLocker’s TPM-only protector is ineffective when the TPM’s internal state can be extracted However, BitLocker’s TPM and PIN protector retains the protection that a PIN-only strategy would offer This suggests that a TPM and PIN protector can be implemented such that the security of a PIN-only strategy remains once the fTPM state is compromised
Discussion on Security Trade-offs
The paper discusses that an FDE implementation must employ standalone anti-brute-force measures beyond the sealed TPM object as BitLocker does A passphrase can provide an adequate level of security even with a compromised fTPM However, a traditional PIN does not provide even minuscule brute-force protection Microsoft should empower their users to make an informed choice regarding the protection level of their data-at-rest
Conclusion
AMD’s firmware TPMs are vulnerable to physical attacks such as voltage fault injection Our work is the first full TPM state compromise against a widely adopted TPM implementation This full state compromise gives the powerful ability to defeat any TPM-based security The paper's arXiv ID is 2304.14717.
Green — the paper says it, and the page is printed. Red — no passage carries it, so treat that sentence as unverified. The same check runs on your own text: check a summary against its source, $5.
Improvements for AI systems
-
Develop a system that can detect physical tampering attempts against firmware TPMs by monitoring voltage fluctuations on buses like SVI2 during boot sequences, as this attack
leverages the Serial Voltage Identification Interface 2.0 (SVI2) bus, allowing the AMD SoC to update its supply voltages dynamically.
-
Create an AI defense mechanism that can analyze the non-volatile storage format of AMD fTPMs, specifically parsing the
NV DATA
file structure to derive keys and decrypt sealed TPM objects using a tool likeamd-nv-tool.
-
Design a security module capable of performing unsealing operations on TPM objects by searching all consecutive 256 bit values in the decrypted non-volatile fTPM state for the correct primary seed, as detailed in Section 4.4.1.
-
Implement an AI that can evaluate the security posture of BitLocker configurations by assessing whether a
TPM and PIN protector retains the protection that a PIN-only strategy would offer,
highlighting the need to avoid relying solely on TPM-only protection for FDE keys. -
Build a system capable of recommending enhanced PINs, such as passphrases, when an fTPM is detected by BitLocker, addressing the finding that
a traditional PIN, i.e., a short numeric secret, does not provide even minuscule brute-force protection.
Abstract
Trusted Platform Modules constitute an integral building block of modern security features. Moreover, as Windows 11 made a TPM 2.0 mandatory, they are subject to an ever-increasing academic challenge. While discrete TPMs - as found in higher-end systems - have been susceptible to attacks on their exposed communication interface, more common firmware TPMs (fTPMs) are immune to this attack vector as they do not communicate with the CPU via an exposed bus. In this paper, we analyze a new class of attacks against fTPMs: Attacking their Trusted Execution Environment can lead to a full TPM state compromise. We experimentally verify this attack by compromising the AMD Secure Processor, which constitutes the TEE for AMD's fTPMs. In contrast to previous dTPM sniffing attacks, this vulnerability exposes the complete internal TPM state of the fTPM. It allows us to extract any cryptographic material stored or sealed by the fTPM regardless of authentication mechanisms such as Platform Configuration Register validation or passphrases with anti-hammering protection. First, we demonstrate the impact of our findings by - to the best of our knowledge - enabling the first attack against Full Disk Encryption solutions backed by an fTPM. Furthermore, we lay out how any application relying solely on the security properties of the TPM - like Bitlocker's TPM- only protector - can be defeated by an attacker with 2-3 hours of physical access to the target device. Lastly, we analyze the impact of our attack on FDE solutions protected by a TPM and PIN strategy. While a naive implementation also leaves the disk completely unprotected, we find that BitLocker's FDE implementation withholds some protection depending on the complexity of the used PIN. Our results show that when an fTPM's internal state is compromised, a TPM and PIN strategy for FDE is less secure than TPM-less protection with a reasonable passphrase.
Related papers
- From Strategy Search to Policy Learning: MCTS-Guided Adversarial Co-Evolution for LLM Safety Alignment
- Privacy Leakage from Recurrent State in LLM Inference
- Incrementally Verifiable Oracle Aided Computation
- Privilege and confidentiality in generative AI workflows
- BadMoE: Backdooring Mixture-of-Experts LLMs via Routing-Aware Triggers and Task-Underutilized Experts
- CoVault: Secure, Scalable Analytics of Personal Data