Inferensys

Glossary

Trusted Platform Module (TPM)

A dedicated microcontroller on a device's motherboard that secures hardware through integrated cryptographic keys, performing integrity measurements and ensuring the boot process has not been tampered with.
Modern WeWork hardware lab area with product team collaborating around AI device prototypes, 3D printer in background, dramatic industrial lighting with product sketches on glass walls.
HARDWARE ROOT OF TRUST

What is Trusted Platform Module (TPM)?

A Trusted Platform Module (TPM) is a dedicated, tamper-resistant microcontroller integrated into a device's motherboard that secures hardware through integrated cryptographic keys, performing integrity measurements and ensuring the boot process has not been tampered with.

A Trusted Platform Module (TPM) is an international standard (ISO/IEC 11889) for a secure cryptoprocessor. It is a physical chip that generates, stores, and limits the use of cryptographic keys. Crucially, the TPM performs measured boot, hashing the firmware and bootloader before execution and storing these integrity measurements in Platform Configuration Registers (PCRs) to detect unauthorized modification.

This hardware root of trust enables remote attestation, allowing a third party to cryptographically verify the exact software stack running on a machine before releasing secrets or granting network access. Unlike a Hardware Security Module (HSM), which is typically an external device for server-scale key management, the TPM is embedded on the motherboard to bind secrets to a specific device identity, ensuring data cannot be decrypted if the drive is removed.

HARDWARE ROOT OF TRUST

Core Capabilities of a TPM

A Trusted Platform Module (TPM) is a passive, tamper-resistant microcontroller that provides a hardware root of trust by generating, storing, and limiting the use of cryptographic keys. It performs critical security functions that are isolated from the host operating system, ensuring integrity and confidentiality even if the OS is compromised.

01

Cryptographic Key Generation & Protection

The TPM contains a true random number generator (TRNG) to create high-entropy asymmetric key pairs directly on the chip. Private keys are sealed within the TPM's shielded location and are never exposed to the system bus, memory, or hard drive. All cryptographic operations using these keys occur inside the TPM's boundary, preventing exfiltration by malware. This capability is the foundation for hardware-backed keystores and offline token generation, ensuring identity credentials cannot be cloned in software.

02

Platform Integrity & Measured Boot

The TPM enforces a measured boot process by acting as a tamper-proof repository for integrity measurements. Before executing any component of the boot chain, the firmware hashes the next component and extends the hash into a Platform Configuration Register (PCR). This creates an immutable, verifiable log of the exact software state. Any modification to the bootloader or OS kernel results in a different PCR value, allowing for remote attestation to detect rootkits or bootkits before secrets are released.

03

Sealed Storage & Data Binding

The TPM can encrypt data and bind it to a specific platform configuration. Using the TPM2_Seal operation, data is encrypted not just with a key, but with a policy that specifies the exact PCR values required for decryption. If the platform state has changed—indicating a compromise or unauthorized software—the TPM will refuse to unseal the data. This is critical for BitLocker Drive Encryption and protecting secrets in air-gapped model deployment scenarios where physical access is the primary threat vector.

04

Remote Attestation

The TPM can produce a signed quote of its current PCR values using an Attestation Identity Key (AIK). A remote verifier can cryptographically validate this quote to confirm the exact software stack running on the machine. This enables Zero Trust Architecture (ZTA) principles by proving a node is healthy before allowing it to join a secure cluster or receive sensitive data. In confidential computing enclaves, this is the mechanism that proves the enclave is running the correct, unmodified code.

05

Platform Configuration Registers (PCRs)

PCRs are the TPM's volatile memory slots that store integrity measurements. They cannot be written to arbitrarily; they are updated exclusively through a cryptographic extend operation: New_PCR = Hash(Old_PCR || New_Hash). This chaining mechanism ensures that the order of events is preserved and that no measurement can be deleted without resetting the platform. PCRs are the core mechanism enabling supply chain integrity verification, as they record every firmware and software component loaded from power-on.

06

Limited-Use Keys & Dictionary Attack Protection

The TPM enforces hardware-level authorization logic that resists brute-force attacks. It implements dictionary attack protection by exponentially increasing the lockout time after repeated failed authorization attempts, a defense impossible to bypass via software manipulation. Furthermore, keys can be created with specific usage constraints, such as requiring a physical presence confirmation or limiting the number of times a key can be used, providing robust defense against automated malware.

HARDWARE SECURITY

Frequently Asked Questions

Explore the critical role of the Trusted Platform Module in establishing a hardware root of trust for sovereign and air-gapped AI infrastructure.

A Trusted Platform Module (TPM) is a dedicated, tamper-resistant microcontroller integrated onto a device's motherboard that secures hardware through integrated cryptographic keys. It functions as a hardware root of trust by generating, storing, and limiting the use of cryptographic keys. The TPM performs integrity measurements during the boot sequence, hashing the firmware, bootloader, and operating system components and storing these values in Platform Configuration Registers (PCRs). This process, known as measured boot, creates an immutable audit log. Because the private key is generated and bound within the silicon and never leaves the chip, it is resistant to software-based extraction attacks, providing a strong foundation for system security.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.