Center for Advanced Studies (CAS) Research Focus Physics and Security

Secure money and payments serve a fundamental societal need: Already over 30,000 years ago, wooden sticks with notches were used as first “payment instruments”, followed by seashells and coins from rare natural materials. With the rise of paper money, sophisticated man-made security features came into focus: Extra elements were added to paper to combat counterfeiting, often operating at the scientific and technological edge of their time. Today, the emergence of electronic payment systems has taken money to its next evolutionary level, including digital currencies, bitcoin, or Apple Pay. Various advanced technologies, such as chip cards, encryption, biometrics, and cryptographic protocols, may interact to secure modern monetary transactions. Can perfect payment security finally be achieved in this manner? The talk puts particular emphasis on the science behind securing money and payments and takes a short “payment journey” through the past, and via Web 3.0 and Metaverse also risks a short glimpse at the future of currencies. | Michael Tagscherer ist Chief Technology Officer von Giesecke & Devrient in München.

In modern electronic systems, analog circuits are primarily used to condition and discretize physical signals, while most of the information processing occurs in the digital domain. The digital abstraction has allowed us to focus on bits as atomic carriers of information, largely ignore analog nuances in the underlying physical signals, and hence scale hierarchical systems toward transistor counts of over 100 billion on a single chip. While this approach has been a key enabler for managing complexity, it has always been scrutinized for its potential inefficiencies. Why don't we prefer to perform an addition by simply connecting two wires carrying analog currents? Similarly, why not compute an integral by collecting charges on a capacitor? This talk will aim to provide some answers to these questions, with the specific focus on domain-specific circuits for machine learning and hardware security. | Boris Murmann is Professor of Electrical Engineering at Stanford University and Fellow in the context of the CAS Research Focus "Physics and Security".

Secure money and payments serve a fundamental societal need: Already over 30,000 years ago, wooden sticks with notches were used as first “payment instruments”, followed by seashells and coins from rare natural materials. With the rise of paper money, sophisticated man-made security features came into focus: Extra elements were added to paper to combat counterfeiting, often operating at the scientific and technological edge of their time. Today, the emergence of electronic payment systems has taken money to its next evolutionary level, including digital currencies, bitcoin, or Apple Pay. Various advanced technologies, such as chip cards, encryption, biometrics, and cryptographic protocols, may interact to secure modern monetary transactions. Can perfect payment security finally be achieved in this manner? The talk puts particular emphasis on the science behind securing money and payments and takes a short “payment journey” through the past, and via Web 3.0 and Metaverse also risks a short glimpse at the future of currencies. | Michael Tagscherer ist Chief Technology Officer von Giesecke & Devrient in München.

In modern electronic systems, analog circuits are primarily used to condition and discretize physical signals, while most of the information processing occurs in the digital domain. The digital abstraction has allowed us to focus on bits as atomic carriers of information, largely ignore analog nuances in the underlying physical signals, and hence scale hierarchical systems toward transistor counts of over 100 billion on a single chip. While this approach has been a key enabler for managing complexity, it has always been scrutinized for its potential inefficiencies. Why don't we prefer to perform an addition by simply connecting two wires carrying analog currents? Similarly, why not compute an integral by collecting charges on a capacitor? This talk will aim to provide some answers to these questions, with the specific focus on domain-specific circuits for machine learning and hardware security. | Boris Murmann is Professor of Electrical Engineering at Stanford University and Fellow in the context of the CAS Research Focus "Physics and Security".

Attackers have targeted software and networking for a long time. Now they are also attacking the hardware foundations of computers. Attacks on microprocessors, the brains of computers, can be devastating. We illustrate with the recent rush of speculative execution attacks that exploit the performance optimization features at the microarchitecture level of modern microprocessors, to leak arbitrary secrets. To defend against these attacks without suffering severe performance degradation and software incompatibility, hardware solutions are desired. In this talk, Lee sheds light on the critical attack steps of speculative execution attacks (also called transient execution attacks), and why these attacks succeed. She discusses potential hardware defense strategies, and defenses proposed by the computer architecture research community. Beyond current attacks and defenses, are there new proactive design strategies that enable future microprocessors to improve both attack-resilience and performance? | Ruby B. Lee is Forest G. Hamrick Professor and Professor of Electrical and Computer Engineering emeritus at Princeton University.

Deep learning (DL) models are enabling a significant paradigm shift in a diverse range of fields, including natural language processing, computer vision, as well as the design and automation of complex integrated circuits. While the deep models – and optimizations- based on them, e.g., Deep Reinforcement Learning (RL) – demonstrate a superior performance and a great capability for automated representation learning, earlier works have revealed the vulnerability of DLs to various attacks. The vulnerabilities include adversarial samples, model poisoning, fault injection, and Intellectual Property (IP) infringement attacks. On the one hand, these security threats could divert the behavior of the DL model and lead to incorrect decisions in critical tasks. On the other hand, the susceptibility of DLs to IP piracy attacks might thwart trustworthy technology transfer as well as reliable DL deployment. In this talk, Farinaz Koushanfar investigates the existing defense techniques to protect DLs against the above-mentioned security threats. Particularly, she reviews end-to-end defense schemes for robust deep learning in both centralized and federated learning settings. Her comprehensive taxonomy and horizontal comparisons reveal an important fact that defense strategies developed using DL/software/hardware co-design outperform the DL/software-only counterparts and show how they can achieve very efficient and latency-optimized defenses for real-world applications. She believes his systemization of knowledge sheds light on the promising performance of hardware-based DL security methodologies and can guide the development of future defenses. | Farinaz Koushanfar is professor in the Electrical and Computer Engineering department at the University of California San Diego (UCSD), where she is also the co-founder and co-director of the UCSD Center for Machine-Intelligence, Computing & Security.

Attackers have targeted software and networking for a long time. Now they are also attacking the hardware foundations of computers. Attacks on microprocessors, the brains of computers, can be devastating. We illustrate with the recent rush of speculative execution attacks that exploit the performance optimization features at the microarchitecture level of modern microprocessors, to leak arbitrary secrets. To defend against these attacks without suffering severe performance degradation and software incompatibility, hardware solutions are desired. In this talk, Lee sheds light on the critical attack steps of speculative execution attacks (also called transient execution attacks), and why these attacks succeed. She discusses potential hardware defense strategies, and defenses proposed by the computer architecture research community. Beyond current attacks and defenses, are there new proactive design strategies that enable future microprocessors to improve both attack-resilience and performance? | Ruby B. Lee is Forest G. Hamrick Professor and Professor of Electrical and Computer Engineering emeritus at Princeton University.

Deep learning (DL) models are enabling a significant paradigm shift in a diverse range of fields, including natural language processing, computer vision, as well as the design and automation of complex integrated circuits. While the deep models – and optimizations- based on them, e.g., Deep Reinforcement Learning (RL) – demonstrate a superior performance and a great capability for automated representation learning, earlier works have revealed the vulnerability of DLs to various attacks. The vulnerabilities include adversarial samples, model poisoning, fault injection, and Intellectual Property (IP) infringement attacks. On the one hand, these security threats could divert the behavior of the DL model and lead to incorrect decisions in critical tasks. On the other hand, the susceptibility of DLs to IP piracy attacks might thwart trustworthy technology transfer as well as reliable DL deployment. In this talk, Farinaz Koushanfar investigates the existing defense techniques to protect DLs against the above-mentioned security threats. Particularly, she reviews end-to-end defense schemes for robust deep learning in both centralized and federated learning settings. Her comprehensive taxonomy and horizontal comparisons reveal an important fact that defense strategies developed using DL/software/hardware co-design outperform the DL/software-only counterparts and show how they can achieve very efficient and latency-optimized defenses for real-world applications. She believes his systemization of knowledge sheds light on the promising performance of hardware-based DL security methodologies and can guide the development of future defenses. | Farinaz Koushanfar is professor in the Electrical and Computer Engineering department at the University of California San Diego (UCSD), where she is also the co-founder and co-director of the UCSD Center for Machine-Intelligence, Computing & Security.