Enforcing network policies over encrypted traffic without requiring significant protocol modifications or violating end-to-end encryption is possible by using the zero-knowledge middlebox (ZKMB) paradigm. In this model, clients attach a zero-knowledge proof (ZKP) to their packets, demonstrating compliance with a network policy without revealing any information about the underlying plaintext. Previous works have explored this architecture and evaluated its practicality primarily in terms of proof-generation latency. However, these solutions do not address scenarios in which the client is a resource-constrained device and therefore cannot efficiently generate the proof on its own. At the same time, several proof-of-concept systems have recently been proposed to either design new protocols or adapt existing ones for delegated and privacy-preserving zero-knowledge proof generation through secure multiparty computation, ensuring that no information about private data is leaked during the process. In this thesis, we adapt one of these algorithms, DFS, to the zero-knowledge middlebox architecture and implement a policy on the HTTPS request body sent by the client. The results show significant advantages in offloading proof computation for resource-constrained clients in terms of CPU usage, memory consumption, and total online time, compared to baseline local proving. Our contributions include a complete architecture and implementation as well as the modifications and improvements required to make the original protocol fit the ZKMB scenario. The results demonstrate the benefits of the proposed architecture, although real-time detection remains infeasible with the current implementation.
L’applicazione di policy di rete su traffico cifrato, senza richiedere modifiche significative ai protocolli né compromettere la cifratura end-to-end, è possibile mediante il paradigma dei middlebox a conoscenza zero (Zero-Knowledge Middleboxes, ZKMB). In questo modello, i client allegano ai propri pacchetti una prova a conoscenza zero (Zero-Knowledge Proof, ZKP), dimostrando la conformità a una policy di rete senza rivelare alcuna informazione sul contenuto del loro traffico. Alcuni lavori precedenti hanno analizzato questa architettura e ne hanno valutato la praticità principalmente in termini di latenza nella generazione della prova. Tuttavia, queste soluzioni non affrontano gli scenari in cui il client è un dispositivo con risorse limitate e non è quindi in grado di generare in modo efficiente la prova in autonomia. Parallelamente, negli ultimi anni sono stati presentati diversi nuovi protocolli (o proposte modifiche a protocolli esistenti) per la generazione delegata e sicura di prove a conoscenza zero tramite calcolo multiparte, garantendo che nessuna informazione relativa ai dati privati venga divulgata durante il processo. In questa tesi, adattiamo uno di questi algoritmi, DFS, all’architettura dei middlebox a conoscenza zero e implementiamo una policy applicata al contenuto delle richieste HTTPS inviate dal client. I risultati mostrano vantaggi significativi nella delegazione della generazione della prova per client con risorse limitate in termini di utilizzo della CPU, consumo di memoria e tempo totale online, rispetto alla generazione locale della prova. I contributi includono un’architettura e un’implementazione completa, nonché le modifiche e i miglioramenti necessari per adattare il protocollo originale allo scenario ZKMB. I risultati dimostrano i benefici dell’architettura proposta, sebbene un rilevamento in tempo reale rimanga ancora irrealizzabile.
A delegation-friendly Zero-Knowledge Middlebox
Motto, Francesco
2025/2026
Abstract
Enforcing network policies over encrypted traffic without requiring significant protocol modifications or violating end-to-end encryption is possible by using the zero-knowledge middlebox (ZKMB) paradigm. In this model, clients attach a zero-knowledge proof (ZKP) to their packets, demonstrating compliance with a network policy without revealing any information about the underlying plaintext. Previous works have explored this architecture and evaluated its practicality primarily in terms of proof-generation latency. However, these solutions do not address scenarios in which the client is a resource-constrained device and therefore cannot efficiently generate the proof on its own. At the same time, several proof-of-concept systems have recently been proposed to either design new protocols or adapt existing ones for delegated and privacy-preserving zero-knowledge proof generation through secure multiparty computation, ensuring that no information about private data is leaked during the process. In this thesis, we adapt one of these algorithms, DFS, to the zero-knowledge middlebox architecture and implement a policy on the HTTPS request body sent by the client. The results show significant advantages in offloading proof computation for resource-constrained clients in terms of CPU usage, memory consumption, and total online time, compared to baseline local proving. Our contributions include a complete architecture and implementation as well as the modifications and improvements required to make the original protocol fit the ZKMB scenario. The results demonstrate the benefits of the proposed architecture, although real-time detection remains infeasible with the current implementation.| File | Dimensione | Formato | |
|---|---|---|---|
|
2026_07_Motto_Tesi.pdf
solo utenti autorizzati a partire dal 01/07/2027
Descrizione: Tesi
Dimensione
1.06 MB
Formato
Adobe PDF
|
1.06 MB | Adobe PDF | Visualizza/Apri |
|
2026_07_Motto_Executive_summary.pdf
solo utenti autorizzati a partire dal 01/07/2027
Descrizione: Executive Summary
Dimensione
576.07 kB
Formato
Adobe PDF
|
576.07 kB | 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/10589/260389