Historically, system security was treated primarily as a software challenge. A robust industry emerged around tools designed to harden applications, detect vulnerabilities, and ensure resilience against cyberattacks. This is the foundation of the DevSecOps ecosystem, where companies such as CrowdStrike, Semperis, and Snyk have built strong businesses providing threat detection, vulnerability management, and secure software development solutions.
While these efforts remain critical to modern cybersecurity, they have largely operated under the assumption that the hardware root of trust—the fundamental building blocks of computing platforms—was inherently secure or prohibitively difficult to compromise. As a result, most security innovation concentrated on the software stack, leaving the underlying hardware layer comparatively underexplored.
Given the advances in understanding hardware related vulnerabilities and their discoveries, and thanks to the exponential rise in sophistication of AI capabilities, compromising hardware is now within reach, and the attack vectors continue to get more aggressive all the time. This creates the need for a new form of DevSecOps that fortifies the firmware and hardware portion of systems. These trends are summarized in the graphic above and the implications are the topic of this post.
How Big is this Problem?
Simply put, it’s very large and growing fast. Over the past five years or so there have been several high-profile security flaws discovered in mainstream hardware. This is the hardware that was part of the immutable root of trust. Generative AI has disrupted a lot of our world. Much of it improves our lives. Here is a case where it can also make life more difficult. And unlike software security, hardware vulnerabilities are not patchable. The battle cry is simply:
Attack one, attack all.
Here are some examples.
Spectre/Meltdown: Intel lost about $12B of its market cap in one day due to attacks attributed to this flaw. The attacks exploited out-of-order execution, speculative execution, and caching resulting from gaps in security considerations. Everyone with a PC or smartphone was at risk. A more recent Intel “Downfall” bug documents similar vulnerability.
GhostWrite: Alibaba’s T-Head was breached by the German Center for Information Security. For the TH1520 SoC, hackers accessed read and write physical memory and executed arbitrary code. Bad actors can take over the SoC and hijack the host. Vulnerability lies in faulty instructions in the chip RTL. Instructions are “baked” into the design in silicon and cannot be fixed with a software update.
ReVault : Is a security bug in a Broadcom chip that puts millions of Dell laptops at risk. A ReVault attack could be used as a physical compromise to bypass Windows Login and/or for any local user to gain Admin/System privilege, allowing a hacker to steal sensitive information and credentials, as well as biometric data like fingerprint information. This one appears to be fixable at the firmware level.
GoFetch: Is associated with Apple M-series chips that contain a vulnerability to a side-channel attack called “GoFetch,” which exploits data memory-dependent prefetchers (DMPs) to extract encryption keys. DMPs are a feature that use memory access patterns to predict which data might be useful, so it can be preloaded into cache memory to speed up access. A group of researchers discovered that the DMP process in Apple M1, M2 and M3 chips could be probed using attacker-selected inputs, and the prefetching behavior analyzed to ultimately predict encryption keys.
There are many more in the news. You get the idea.
Solutions
In modern system-on-chip (SoC) security, solutions can broadly be categorized into two complementary approaches:
1. Security IP (Functional Security Blocks):
These are dedicated hardware intellectual property (IP) blocks that provide security services within the SoC. Typical examples include:
- - Cryptographic accelerators for encryption, decryption, and hashing
- - True random number generators (TRNGs) for entropy generation
- - Physical unclonable functions (PUFs) for device-unique key generation
- - Secure key storage and memory protection units
- - Trusted Execution Environments (TEEs) and isolated enclaves
2. Secure IP (Asset Protection and Verification):
This category refers to ensuring that existing SoC IP blocks and the SoC as a whole operate securely, with protection of assets such as cryptographic keys, firmware, and sensitive user data. For example:
- - A CPU core may handle sensitive application data and must be verified against side-channel leakage, fault injection vulnerabilities, and privilege escalation risks.
- - A crypto accelerator must ensure confidentiality and integrity of key material, preventing unauthorized access or modification.
- - At the SoC integration level, interconnects, buses, and memory subsystems must be validated to ensure secure data transfer and isolation between trusted and untrusted domains.
In summary:
- - Security IP provides the functional mechanisms that implement security features within an SoC.
- - Secure IP ensures that all assets within an IP block or SoC are protected, maintaining confidentiality, integrity, and availability.
Do Secure Crypto Processors Help?
All forms of security hardening are welcome addition in the face of exponentially growing security threats. The key is to trust but verify. Using security IP DOES NOT necessarily result in a secure system.
The focus of secure crypto processors is to perform cryptographic operations with high security. These operations can have a high yield in terms of data compromise. Unfortunately, all approaches are not completely invulnerable. Sophisticated actors, including government intelligence agencies, can exploit vulnerabilities with sufficient investment in resources. Potential areas of weakness include:
- - Physical Attacks: Vulnerable to direct tampering and extraction of data.
- - Hardware Backdoors: Design flaws can allow unauthorized access.
- - Cold Boot Attacks: Data remanence can be exploited to access sensitive information.
- - Weak Key Management: Poor practices can compromise security.
- - Insecure Random Generation: Flawed randomness can weaken cryptographic processes.
- - Side-Channel Attacks: Methods to decode information leaked during cryptographic operations, such as timing information or power consumption.
- - Software Exploits: Access of decrypted data or keys if the security envelope is compromised.
If you’d like to learn more about some of these scenarios, Microsoft has published a detailed research report here. This report deals with cryptojacking – a type of cyberattack that uses stolen computing power to mine cryptocurrency. A broader view of cryptographic failures with advanced insights and descriptions of real-world attacks can be found in this post. Topics discussed include:
- - Side-Channel Attacks
- - Failure to Use Authenticated Encryption
- - Poor Key Exchange Mechanisms
- - Deprecated Digital Signatures
The National Institute of Standards and Technology (NIST) is driving a lot of this work with enhanced requirements, creating the post-quantum cryptography (PQC) movement. Here again, there is always room for better verification and hardening. A while ago a news piece reported that post-quantum crypto was cracked in an hour with one core of an ancient Xeon processor.
What’s Next?
Higher levels of security are critical to continued advances of AI technologies. There are many approaches that can be used to accomplish this goal, but a strategy of trust but verify is very relevant. There is a difference between security IP and secure IP, and that difference could easily be the margin of victory for your next project.
The good news is there is a company that focuses exclusively on the security verification of chip IP. The technology fits into existing design flows and uses generative AI to find and help fix security flaws early in the design process. This approach should be applied to every design and every piece of IP, regardless of the stated security credentials. Trust but verify.
The company is Caspia Technologies. Caspia’s founding team brings together over 75 years of experience in various fields of the semiconductor market, including design, fabrication, test, EDA development, and importantly security and trust. Caspia delivers security-focused solutions to both public and private customers to enhance electronic designs and microelectronics physical hardware assurance. The company has pioneered the new approach to hardware DevSecOps shown in the graphic at the top of this post.
You can learn more about this unique and important company here. If hardware security worries you (and it should), you can reach out to set up a meeting with Caspia here to begin your journey to a safer tomorrow. And that’s how you can tell the difference between security IP and secure IP.
Link to Press Release