FreshRSS

Zobrazení pro čtení

Jsou dostupné nové články, klikněte pro obnovení stránky.

Securing AI In The Data Center

AI has permeated virtually every aspect of our digital lives, from personalized recommendations on streaming platforms to advanced medical diagnostics. Behind the scenes of this AI revolution lies the data center, which houses the hardware, software, and networking infrastructure necessary for training and deploying AI models. Securing AI in the data center relies on data confidentiality, integrity, and authenticity throughout the AI lifecycle, from data preprocessing to model training and inference deployment.

High-value datasets containing sensitive information, such as personal health records or financial transactions, must be shielded from unauthorized access. Robust encryption mechanisms, such as Advanced Encryption Standard (AES), coupled with secure key management practices, form the foundation of data confidentiality in the data center. The encryption key used must be unique and used in a secure environment. Encryption and decryption operations of data are constantly occurring and must be performed to prevent key leakage. Should a compromise arise, it should be possible to renew the key securely and re-encrypt data with the new key.

The encryption key used must also be securely stored in a location that unauthorized processes or individuals cannot access. The keys used must be protected from attempts to read them from the device or attempts to steal them using side-channel techniques such as SCA (Side-Channel Attacks) or FIA (Fault Injection Attacks). The multitenancy aspect of modern data centers calls for robust SCA protection of key data.

Hardware-level security plays a pivotal role in safeguarding AI within the data center, offering built-in protections against a wide range of threats. Trusted Platform Modules (TPMs), secure enclaves, and Hardware Security Modules (HSMs) provide secure storage and processing environments for sensitive data and cryptographic keys, shielding them from unauthorized access or tampering. By leveraging hardware-based security features, organizations can enhance the resilience of their AI infrastructure and mitigate the risk of attacks targeting software vulnerabilities.

Ideally, secure cryptographic processing is handled by a Root of Trust core. The AI service provider manages the Root of Trust firmware, but it can also load secure applications that customers can write to implement their own cryptographic key management and storage applications. The Root of Trust can be integrated in the host CPU that orchestrates the AI operations, decrypting the AI model and its specific parameters before those are fed to AI or network accelerators (GPUs or NPUs). It can also be directly integrated with the GPUs and NPUs to perform encryption/decryption at that level. These GPUs and NPUs may also select to store AI workloads and inference models in encrypted form in their local memory banks and decrypt the data on the fly when access is required. Dedicated on-the-fly, low latency in-line memory decryption engines based on the AES-XTS algorithm can keep up with the memory bandwidth, ensuring that the process is not slowed down.

AI training workloads are often distributed among dozens of devices connected via PCIe or high-speed networking technology such as 800G Ethernet. An efficient confidentiality and integrity protocol such as MACsec using the AES-GCM algorithm can protect the data in motion over high-speed Ethernet links. AES-GCM engines integrated with the server SoC and the PCIe acceleration boards ensure that traffic is authenticated and optionally encrypted.

Rambus offers a broad portfolio of security IP covering the key security elements needed to protect AI in the data center. Rambus Root of Trust IP cores ensure a secure boot protocol that protects the integrity of its firmware. This can be combined with Rambus inline memory encryption engines, as well as dedicated solutions for MACsec up to 800G.

Resources

The post Securing AI In The Data Center appeared first on Semiconductor Engineering.

Earning Digital Trust

The internet of things (IoT) has been growing at a fast pace. In 2023, there were already double the number of internet-connected devices – 16 billion – than people on the planet. However, many of these devices are not properly secured. The high volume of insecure devices being deployed is presenting hackers with more opportunities than ever before. Governments around the world are realizing that additional security standards for IoT devices are needed to address the growing and important role of the billions of connected devices we rely on every day. The EU Cyber Resilience Act and the IoT Cybersecurity Improvement Act in the United States are driving improved security practices as well as an increased sense of urgency.

Digital trust is critical for the continued success of the IoT. This means that security, privacy, and reliability are becoming top concerns. IoT devices are always connected and can be deployed in any environment, which means that they can be attacked via the internet as well as physically in the field. Whether it is a remote attacker getting access to a baby monitor or camera inside your house, or someone physically tampering with sensors that are part of a critical infrastructure, IoT devices need to have proper security in place.

This is even more salient when one considers that each IoT device is part of a multi-party supply chain and is used in systems that contain many other devices. All these devices need to be trusted and communicate in a secure way to maintain the privacy of their data. It is critical to ensure that there are no backdoors left open by any link in the supply chain, or when devices are updated in the field. Any weak link exposes more than just the device in question to security breaches; it exposes its entire system – and the IoT itself – to attacks.

A foundation of trust starts in the hardware

To secure the IoT, each piece of silicon in the supply chain needs to be trusted. The best way to achieve this is by using a hardware-based root of trust (RoT) for every device. An RoT is typically defined as “the set of implicitly trusted functions that the rest of the system or device can use to ensure security.” The core of an RoT consists of an identity and cryptographic keys rooted in the hardware of a device. This establishes a unique, immutable, and unclonable identity to authenticate a device in the IoT network. It establishes the anchor point for the chain of trust, and powers critical system security use cases over the entire lifecycle of a device.

Protecting every device on the IoT with a hardware-based RoT can appear to be an unreachable goal. There are so many types of systems and devices and so many different semiconductor and device manufacturers, each with their own complex supply chain. Many of these chips and devices are high-volume/low-cost and therefore have strict constraints on additional manufacturing or supply chain costs for security. The PSA Certified 2023 Security Report indicates that 72% of tech decision makers are interested in the development of an industry-led set of guidelines to make reaching the goal of a secure IoT more attainable.

Security frameworks and certifications speed-up the process and build confidence

One important industry-led effort in standardizing IoT security that has been widely adopted is PSA Certified. PSA stands for Platform Security Architecture and PSA Certified is a global partnership addressing security challenges and uniting the technology ecosystem under a common security baseline, providing an easy-to consume and comprehensive methodology for the lab-validated assurance of device security. PSA Certified has been adopted by the full supply chain from silicon providers, software vendors, original equipment manufacturers (OEMs), IP providers, governments, content service providers (CSPs), insurance vendors and other third-party schemes. PSA Certified was the winner of the IoT Global Awards “Ecosystem of the year” in 2021.

PSA Certified lab-based evaluations (PSA Certified Level 2 and above) have a choice of evaluation methodologies, including the rigorous SESIP-based methodology (Security Evaluation Standard for IoT Platforms from GlobalPlatform), an optimized security evaluation methodology, designed for connected devices. PSA Certified recognizes that a myriad of different regulations and certification frameworks create an added layer of complexity for the silicon providers, OEMs, software vendors, developers, and service providers tasked with demonstrating the security capability of their products. The goal of the program is to provide a flexible and efficient security evaluation method needed to address the unique complexities and challenges of the evolving digital ecosystem and to drive consistency across device certification schemes to bring greater trust.

The PSA Certified framework recognizes the importance of a hardware RoT for every connected device. It currently provides incremental levels of certified assurance, ranging from a baseline Level 1 (application of best-practice security principles) to a more advanced Level 3 (validated protection against substantial hardware and software attacks).

PSA Certified RoT component

Among the certifications available, PSA Certified offers a PSA Certified RoT Component certification program, which targets separate RoT IP components, such as physical unclonable functions (PUFs), which use unclonable properties of silicon to create a robust trust (or security) anchor. As shown in figure 1, the PSA-RoT Certification includes three levels of security testing. These component-level certifications from PSA Certified validate specific security functional requirements (SFRs) provided by an RoT component and enable their reuse in a fast-track evaluation of a system integration using this component.

Fig. 1: PSA Certified establishes a chain of trust that begins with a PSA-RoT.

A proven RoT IP solution, now PSA Certified

Synopsys PUF IP is a secure key generation and storage solution that enables device manufacturers and designers to secure their products with internally generated unclonable identities and device-unique cryptographic keys. It uses the inherently random start-up values of SRAM as a physical unclonable function (PUF), which generates the entropy required for a strong hardware root of trust.

This root key created by Synopsys PUF IP is never stored, but rather recreated from the PUF upon each use, so there is never a key to be discovered by attackers. The root key is the basis for key management capabilities that enable each member of the supply chain to create its own secret keys, bound to the specific device, to protect their IP/communications without revealing these keys to any other member of the supply chain.

Synopsys PUF IP offers robust PUF-based physical security, with the following properties:

  • No secrets/keys at rest​ (no secrets stored in any memory)
    • prevents any attack on an unpowered device​
    • keys are only present when used​, limiting the window of opportunity for attacks
  • Hardware entropy source/root of trust​
    • no dependence on third parties​ (no key injection from outside)
    • no dependence on security of external components or other internal modules​
    • no dependence on software-based security​
  • Technology-independent, fully digital standard-logic CMOS IP
    • all fabs and technology nodes
    • small footprint
    • re-use in new platforms/deployments
  • Built-in error resilience​ due to advanced error-correction

The Synopsys PUF technology has been field-proven over more than a decade of deployment on over 750 million chips. And now, the Synopsys PUF has achieved the milestone of becoming the world’s first IP solution to be awarded “PSA Certified Level 3 RoT Component.” This certifies that the IP includes substantial protection against both software and hardware attacks (including side-channel and fault injection attacks) and is qualified as a trusted component in a system that requires PSA Level 3 certification.

Fault detection and other countermeasures

In addition to its PUF-related protection against physical attacks, all Synopsys PUF IP products have several built-in physical countermeasures. These include both systemic security features (such as data format validation, data authentication, key use restrictions, built in self-tests (BIST), and heath checks) as well as more specific countermeasures (such as data masking and dummy cycles) that protect against specific attacks.

The PSA Certified Synopsys PUF IP goes even one step further. It validates all inputs through integrity checks and error detection. It continuously asserts that everything runs as intended, flags any observed faults, and ensures security. Additionally, the PSA Certified Synopsys PUF IP provides hardware and software handholds to the user which assist in checking that all data is correctly transferred into and out of the PUF IP. The Synopsys PUF IP driver also supports fault detection and reporting.

Advantages of PUFs over traditional key injection and storage methods

For end-product developers, PUF IP has many advantages over traditional approaches for key management. These traditional approaches typically require key injection (provisioning secret keys into a device) and some form of non-volatile memory (NVM), such as embedded Flash memory or one-time programmable storage (OTP), where the programmed key is stored and where it needs to be protected from being extracted, overwritten, or changed. Unlike these traditional key injection solutions, Synopsys PUF IP does not require sensitive key handling by third parties, since PUF-based keys are created within the device itself. In addition, Synopsys PUF IP offers more flexibility than traditional solutions, as a virtually unlimited number of PUF-based keys can be created. And keys protected by the PUF can be added at any time in the lifecycle rather than only during manufacturing.

In terms of key storage, Synopsys PUF IP offers higher protection against physical attacks than storing keys in some form of NVM. PUF-based root keys are not stored on the device, but they are reconstructed upon each use, so there is nothing for attackers to find on the chip. Instead of storing keys in NVM, Synopsys PUF IP stores only (non-sensitive) helper data and encrypted keys in NVM on- or off-chip. The traditional approach of storing keys on the device in NVM is more vulnerable to physical attacks.

Finally, Synopsys PUF IP provides more portability. Since the Synopsys PUF IP is based on standard SRAM memory cells, it offers a process- and fab agnostic solution for key storage that scales to the most advanced technology nodes.

Conclusion

The large and steady increase in devices connected to the IoT also increases the need for digital trust and privacy. This requires flexible and efficient IoT security solutions that are standardized to streamline implementation and certification across the multiple players involved in the creation and deployment of IoT devices. The PSA Certified framework offers an easy-to-consume and comprehensive methodology for the lab-validated assurance of device security.

Synopsys PUF IP, which has been deployed in over 750 million chips, is the first-ever IP solution to be awarded “PSA Certified Level 3 RoT Component.” This certifies that the IP includes substantial protection against hardware and software attacks. Synopsys PUF IP offers IoT device makers a robust PUF-based security anchor with trusted industry-standard certification and offers the perfect balance between strong security, high flexibility, and low cost.

The post Earning Digital Trust appeared first on Semiconductor Engineering.

Securing AI In The Data Center

AI has permeated virtually every aspect of our digital lives, from personalized recommendations on streaming platforms to advanced medical diagnostics. Behind the scenes of this AI revolution lies the data center, which houses the hardware, software, and networking infrastructure necessary for training and deploying AI models. Securing AI in the data center relies on data confidentiality, integrity, and authenticity throughout the AI lifecycle, from data preprocessing to model training and inference deployment.

High-value datasets containing sensitive information, such as personal health records or financial transactions, must be shielded from unauthorized access. Robust encryption mechanisms, such as Advanced Encryption Standard (AES), coupled with secure key management practices, form the foundation of data confidentiality in the data center. The encryption key used must be unique and used in a secure environment. Encryption and decryption operations of data are constantly occurring and must be performed to prevent key leakage. Should a compromise arise, it should be possible to renew the key securely and re-encrypt data with the new key.

The encryption key used must also be securely stored in a location that unauthorized processes or individuals cannot access. The keys used must be protected from attempts to read them from the device or attempts to steal them using side-channel techniques such as SCA (Side-Channel Attacks) or FIA (Fault Injection Attacks). The multitenancy aspect of modern data centers calls for robust SCA protection of key data.

Hardware-level security plays a pivotal role in safeguarding AI within the data center, offering built-in protections against a wide range of threats. Trusted Platform Modules (TPMs), secure enclaves, and Hardware Security Modules (HSMs) provide secure storage and processing environments for sensitive data and cryptographic keys, shielding them from unauthorized access or tampering. By leveraging hardware-based security features, organizations can enhance the resilience of their AI infrastructure and mitigate the risk of attacks targeting software vulnerabilities.

Ideally, secure cryptographic processing is handled by a Root of Trust core. The AI service provider manages the Root of Trust firmware, but it can also load secure applications that customers can write to implement their own cryptographic key management and storage applications. The Root of Trust can be integrated in the host CPU that orchestrates the AI operations, decrypting the AI model and its specific parameters before those are fed to AI or network accelerators (GPUs or NPUs). It can also be directly integrated with the GPUs and NPUs to perform encryption/decryption at that level. These GPUs and NPUs may also select to store AI workloads and inference models in encrypted form in their local memory banks and decrypt the data on the fly when access is required. Dedicated on-the-fly, low latency in-line memory decryption engines based on the AES-XTS algorithm can keep up with the memory bandwidth, ensuring that the process is not slowed down.

AI training workloads are often distributed among dozens of devices connected via PCIe or high-speed networking technology such as 800G Ethernet. An efficient confidentiality and integrity protocol such as MACsec using the AES-GCM algorithm can protect the data in motion over high-speed Ethernet links. AES-GCM engines integrated with the server SoC and the PCIe acceleration boards ensure that traffic is authenticated and optionally encrypted.

Rambus offers a broad portfolio of security IP covering the key security elements needed to protect AI in the data center. Rambus Root of Trust IP cores ensure a secure boot protocol that protects the integrity of its firmware. This can be combined with Rambus inline memory encryption engines, as well as dedicated solutions for MACsec up to 800G.

Resources

The post Securing AI In The Data Center appeared first on Semiconductor Engineering.

Earning Digital Trust

The internet of things (IoT) has been growing at a fast pace. In 2023, there were already double the number of internet-connected devices – 16 billion – than people on the planet. However, many of these devices are not properly secured. The high volume of insecure devices being deployed is presenting hackers with more opportunities than ever before. Governments around the world are realizing that additional security standards for IoT devices are needed to address the growing and important role of the billions of connected devices we rely on every day. The EU Cyber Resilience Act and the IoT Cybersecurity Improvement Act in the United States are driving improved security practices as well as an increased sense of urgency.

Digital trust is critical for the continued success of the IoT. This means that security, privacy, and reliability are becoming top concerns. IoT devices are always connected and can be deployed in any environment, which means that they can be attacked via the internet as well as physically in the field. Whether it is a remote attacker getting access to a baby monitor or camera inside your house, or someone physically tampering with sensors that are part of a critical infrastructure, IoT devices need to have proper security in place.

This is even more salient when one considers that each IoT device is part of a multi-party supply chain and is used in systems that contain many other devices. All these devices need to be trusted and communicate in a secure way to maintain the privacy of their data. It is critical to ensure that there are no backdoors left open by any link in the supply chain, or when devices are updated in the field. Any weak link exposes more than just the device in question to security breaches; it exposes its entire system – and the IoT itself – to attacks.

A foundation of trust starts in the hardware

To secure the IoT, each piece of silicon in the supply chain needs to be trusted. The best way to achieve this is by using a hardware-based root of trust (RoT) for every device. An RoT is typically defined as “the set of implicitly trusted functions that the rest of the system or device can use to ensure security.” The core of an RoT consists of an identity and cryptographic keys rooted in the hardware of a device. This establishes a unique, immutable, and unclonable identity to authenticate a device in the IoT network. It establishes the anchor point for the chain of trust, and powers critical system security use cases over the entire lifecycle of a device.

Protecting every device on the IoT with a hardware-based RoT can appear to be an unreachable goal. There are so many types of systems and devices and so many different semiconductor and device manufacturers, each with their own complex supply chain. Many of these chips and devices are high-volume/low-cost and therefore have strict constraints on additional manufacturing or supply chain costs for security. The PSA Certified 2023 Security Report indicates that 72% of tech decision makers are interested in the development of an industry-led set of guidelines to make reaching the goal of a secure IoT more attainable.

Security frameworks and certifications speed-up the process and build confidence

One important industry-led effort in standardizing IoT security that has been widely adopted is PSA Certified. PSA stands for Platform Security Architecture and PSA Certified is a global partnership addressing security challenges and uniting the technology ecosystem under a common security baseline, providing an easy-to consume and comprehensive methodology for the lab-validated assurance of device security. PSA Certified has been adopted by the full supply chain from silicon providers, software vendors, original equipment manufacturers (OEMs), IP providers, governments, content service providers (CSPs), insurance vendors and other third-party schemes. PSA Certified was the winner of the IoT Global Awards “Ecosystem of the year” in 2021.

PSA Certified lab-based evaluations (PSA Certified Level 2 and above) have a choice of evaluation methodologies, including the rigorous SESIP-based methodology (Security Evaluation Standard for IoT Platforms from GlobalPlatform), an optimized security evaluation methodology, designed for connected devices. PSA Certified recognizes that a myriad of different regulations and certification frameworks create an added layer of complexity for the silicon providers, OEMs, software vendors, developers, and service providers tasked with demonstrating the security capability of their products. The goal of the program is to provide a flexible and efficient security evaluation method needed to address the unique complexities and challenges of the evolving digital ecosystem and to drive consistency across device certification schemes to bring greater trust.

The PSA Certified framework recognizes the importance of a hardware RoT for every connected device. It currently provides incremental levels of certified assurance, ranging from a baseline Level 1 (application of best-practice security principles) to a more advanced Level 3 (validated protection against substantial hardware and software attacks).

PSA Certified RoT component

Among the certifications available, PSA Certified offers a PSA Certified RoT Component certification program, which targets separate RoT IP components, such as physical unclonable functions (PUFs), which use unclonable properties of silicon to create a robust trust (or security) anchor. As shown in figure 1, the PSA-RoT Certification includes three levels of security testing. These component-level certifications from PSA Certified validate specific security functional requirements (SFRs) provided by an RoT component and enable their reuse in a fast-track evaluation of a system integration using this component.

Fig. 1: PSA Certified establishes a chain of trust that begins with a PSA-RoT.

A proven RoT IP solution, now PSA Certified

Synopsys PUF IP is a secure key generation and storage solution that enables device manufacturers and designers to secure their products with internally generated unclonable identities and device-unique cryptographic keys. It uses the inherently random start-up values of SRAM as a physical unclonable function (PUF), which generates the entropy required for a strong hardware root of trust.

This root key created by Synopsys PUF IP is never stored, but rather recreated from the PUF upon each use, so there is never a key to be discovered by attackers. The root key is the basis for key management capabilities that enable each member of the supply chain to create its own secret keys, bound to the specific device, to protect their IP/communications without revealing these keys to any other member of the supply chain.

Synopsys PUF IP offers robust PUF-based physical security, with the following properties:

  • No secrets/keys at rest​ (no secrets stored in any memory)
    • prevents any attack on an unpowered device​
    • keys are only present when used​, limiting the window of opportunity for attacks
  • Hardware entropy source/root of trust​
    • no dependence on third parties​ (no key injection from outside)
    • no dependence on security of external components or other internal modules​
    • no dependence on software-based security​
  • Technology-independent, fully digital standard-logic CMOS IP
    • all fabs and technology nodes
    • small footprint
    • re-use in new platforms/deployments
  • Built-in error resilience​ due to advanced error-correction

The Synopsys PUF technology has been field-proven over more than a decade of deployment on over 750 million chips. And now, the Synopsys PUF has achieved the milestone of becoming the world’s first IP solution to be awarded “PSA Certified Level 3 RoT Component.” This certifies that the IP includes substantial protection against both software and hardware attacks (including side-channel and fault injection attacks) and is qualified as a trusted component in a system that requires PSA Level 3 certification.

Fault detection and other countermeasures

In addition to its PUF-related protection against physical attacks, all Synopsys PUF IP products have several built-in physical countermeasures. These include both systemic security features (such as data format validation, data authentication, key use restrictions, built in self-tests (BIST), and heath checks) as well as more specific countermeasures (such as data masking and dummy cycles) that protect against specific attacks.

The PSA Certified Synopsys PUF IP goes even one step further. It validates all inputs through integrity checks and error detection. It continuously asserts that everything runs as intended, flags any observed faults, and ensures security. Additionally, the PSA Certified Synopsys PUF IP provides hardware and software handholds to the user which assist in checking that all data is correctly transferred into and out of the PUF IP. The Synopsys PUF IP driver also supports fault detection and reporting.

Advantages of PUFs over traditional key injection and storage methods

For end-product developers, PUF IP has many advantages over traditional approaches for key management. These traditional approaches typically require key injection (provisioning secret keys into a device) and some form of non-volatile memory (NVM), such as embedded Flash memory or one-time programmable storage (OTP), where the programmed key is stored and where it needs to be protected from being extracted, overwritten, or changed. Unlike these traditional key injection solutions, Synopsys PUF IP does not require sensitive key handling by third parties, since PUF-based keys are created within the device itself. In addition, Synopsys PUF IP offers more flexibility than traditional solutions, as a virtually unlimited number of PUF-based keys can be created. And keys protected by the PUF can be added at any time in the lifecycle rather than only during manufacturing.

In terms of key storage, Synopsys PUF IP offers higher protection against physical attacks than storing keys in some form of NVM. PUF-based root keys are not stored on the device, but they are reconstructed upon each use, so there is nothing for attackers to find on the chip. Instead of storing keys in NVM, Synopsys PUF IP stores only (non-sensitive) helper data and encrypted keys in NVM on- or off-chip. The traditional approach of storing keys on the device in NVM is more vulnerable to physical attacks.

Finally, Synopsys PUF IP provides more portability. Since the Synopsys PUF IP is based on standard SRAM memory cells, it offers a process- and fab agnostic solution for key storage that scales to the most advanced technology nodes.

Conclusion

The large and steady increase in devices connected to the IoT also increases the need for digital trust and privacy. This requires flexible and efficient IoT security solutions that are standardized to streamline implementation and certification across the multiple players involved in the creation and deployment of IoT devices. The PSA Certified framework offers an easy-to-consume and comprehensive methodology for the lab-validated assurance of device security.

Synopsys PUF IP, which has been deployed in over 750 million chips, is the first-ever IP solution to be awarded “PSA Certified Level 3 RoT Component.” This certifies that the IP includes substantial protection against hardware and software attacks. Synopsys PUF IP offers IoT device makers a robust PUF-based security anchor with trusted industry-standard certification and offers the perfect balance between strong security, high flexibility, and low cost.

The post Earning Digital Trust appeared first on Semiconductor Engineering.

❌