In recent decades, the microelectronics industry has witnessed exponential growth in performance and computing capabilities, driven by the relentless scaling of electronic components, resulting in an increasing number of transistors on integrated circuits (IC). However, this scaling trajectory is reaching its limits due to structural and physical constraints, such as the heat wall and the consequent ceiling imposed on clock frequency. Additionally, conventional computing systems based on the von Neumann architecture face inherent challenges stemming from the separation of processing and memory units, leading to significant performance disparities known as the memory wall. This architectural setup proves inefficient, particularly in applications requiring extensive data processing, such as machine learning, due to the sluggish data transfer rates between the CPU and memory. To address these constraints, especially in the era of the Internet of Things (IoT) and Big Data, interest has surged in alternative computing paradigms like in-memory computing (IMC), neuromorphic computing, and stochastic computing. In this context, emerging memory technologies, notably memristors such as resistive switching random access memory (RRAM), phase change memory (PCM), ferroelectric memory (FeRAM), and spin-transfer torque magnetic memory (STT-MRAM), are being explored for their non-volatility, scalability, low power consumption, and fast operation, as well as their compatibility with the complementary metal oxide semiconductor (CMOS) process. However, leveraging these devices in practical applications requires addressing the challenges associated with their stochastic nature. On the other hand, it is precisely the inherent stochasticity of these devices that becomes a strong point in favor of their use in security-related applications. This doctoral thesis aims to explore the development of Physical Unclonable Functions (PUFs) based on emerging non-volatile memory (NVM) technologies. In the realm of hardware security, PUFs can provide a unique physical fingerprint to devices in the Internet of Things (IoT), which is a valuable means of enhancing security through the generation of unique and volatile cryptographic keys with no need to store them in non-volatile memory. By capitalizing on the inherent stochastic characteristics of emerging NVM devices, this research work seeks to develop PUFs that offer enhanced security features while addressing the limitations associated with their use in practical applications and proposing solutions to mitigate them. Through a comprehensive analysis, extensive physics-based simulations, and experimental validations, the dissertation contributes to advancing the understanding and utilization of emerging NVM technologies for secure hardware authentication and cryptographic applications.
Negli ultimi decenni, l'industria microelettronica ha assistito a una crescita esponenziale delle prestazioni e delle capacità di calcolo, guidata dall'inarrestabile scalabilità dei componenti elettronici, che ha portato a un numero crescente di transistor sui circuiti integrati (IC). Tuttavia, questa traiettoria di scalabilità sta raggiungendo i suoi limiti a causa di vincoli strutturali e fisici, come il muro di calore e il conseguente tetto imposto alla frequenza di clock. Inoltre, i sistemi di elaborazione convenzionali basati sull'architettura di von Neumann devono affrontare problemi intrinseci derivanti dalla separazione delle unità di elaborazione e di memoria, che portano a significative disparità di prestazioni, note come "memory wall". Questa configurazione architettonica si rivela inefficiente, in particolare nelle applicazioni che richiedono un'ampia elaborazione dei dati, come l'apprendimento automatico, a causa della lentezza della velocità di trasferimento dei dati tra la CPU e la memoria. Per affrontare questi vincoli, soprattutto nell'era dell'Internet delle cose (IoT) e dei Big Data, è cresciuto l'interesse per paradigmi di calcolo alternativi come l'in-memory computing (IMC), il neuromorphic computing e lo stochastic computing. In questo contesto, le tecnologie di memoria emergenti, in particolare i memristori come le memorie a commutazione resistiva (RRAM), le memorie a cambiamento di fase (PCM), le memorie ferroelettriche (FeRAM) e le memorie magnetiche a coppia di spin-transfer (STT-MRAM), vengono studiate per la loro non volatilità, scalabilità, basso consumo energetico e rapidità di funzionamento, oltre che per la loro compatibilità con il processo di semiconduttori complementari a ossidi metallici (CMOS). Tuttavia, per sfruttare questi dispositivi in applicazioni pratiche è necessario affrontare le sfide associate alla loro natura stocastica. D'altra parte, è proprio la stocasticità intrinseca di questi dispositivi che diventa un punto di forza a favore del loro utilizzo in applicazioni legate alla sicurezza. Questa tesi di dottorato si propone l'obiettivo di esplorare lo sviluppo di funzioni fisiche non clonabili (PUF) basate sulle tecnologie emergenti di memoria non volatile (NVM). Nell'ambito della sicurezza hardware, le PUF possono fornire un'impronta digitale fisica univoca ai dispositivi dell'Internet degli oggetti (IoT), un mezzo prezioso per migliorare la sicurezza attraverso la generazione di chiavi crittografiche univoche e volatili senza la necessità di memorizzarle nella memoria non volatile. Sfruttando le caratteristiche stocastiche intrinseche dei dispositivi NVM emergenti, questo lavoro di ricerca si focalizza sullo sviluppo di dispositivi PUF che offrano maggiori caratteristiche di sicurezza, affrontando al contempo le limitazioni associate al loro utilizzo nelle applicazioni pratiche e proponendo soluzioni per mitigarle. Attraverso un'analisi completa, ampie simulazioni basate sulla fisica e convalide sperimentali, la tesi contribuisce a far progredire la comprensione e l'utilizzo delle tecnologie NVM emergenti per applicazioni sicure di autenticazione e crittografia hardware.
Design and evaluation of memristor-memory-based strong and weak physical unclonable functions
Cattaneo, Lorenzo
2023/2024
Abstract
In recent decades, the microelectronics industry has witnessed exponential growth in performance and computing capabilities, driven by the relentless scaling of electronic components, resulting in an increasing number of transistors on integrated circuits (IC). However, this scaling trajectory is reaching its limits due to structural and physical constraints, such as the heat wall and the consequent ceiling imposed on clock frequency. Additionally, conventional computing systems based on the von Neumann architecture face inherent challenges stemming from the separation of processing and memory units, leading to significant performance disparities known as the memory wall. This architectural setup proves inefficient, particularly in applications requiring extensive data processing, such as machine learning, due to the sluggish data transfer rates between the CPU and memory. To address these constraints, especially in the era of the Internet of Things (IoT) and Big Data, interest has surged in alternative computing paradigms like in-memory computing (IMC), neuromorphic computing, and stochastic computing. In this context, emerging memory technologies, notably memristors such as resistive switching random access memory (RRAM), phase change memory (PCM), ferroelectric memory (FeRAM), and spin-transfer torque magnetic memory (STT-MRAM), are being explored for their non-volatility, scalability, low power consumption, and fast operation, as well as their compatibility with the complementary metal oxide semiconductor (CMOS) process. However, leveraging these devices in practical applications requires addressing the challenges associated with their stochastic nature. On the other hand, it is precisely the inherent stochasticity of these devices that becomes a strong point in favor of their use in security-related applications. This doctoral thesis aims to explore the development of Physical Unclonable Functions (PUFs) based on emerging non-volatile memory (NVM) technologies. In the realm of hardware security, PUFs can provide a unique physical fingerprint to devices in the Internet of Things (IoT), which is a valuable means of enhancing security through the generation of unique and volatile cryptographic keys with no need to store them in non-volatile memory. By capitalizing on the inherent stochastic characteristics of emerging NVM devices, this research work seeks to develop PUFs that offer enhanced security features while addressing the limitations associated with their use in practical applications and proposing solutions to mitigate them. Through a comprehensive analysis, extensive physics-based simulations, and experimental validations, the dissertation contributes to advancing the understanding and utilization of emerging NVM technologies for secure hardware authentication and cryptographic applications.File | Dimensione | Formato | |
---|---|---|---|
Cattaneo_Lorenzo_PhD_Thesis_11072024.pdf
accessibile in internet solo dagli utenti autorizzati
Descrizione: PhD Thesis, PDF
Dimensione
22.08 MB
Formato
Adobe PDF
|
22.08 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/224472