UCLA

5G and 5G IoT technologies have been widely deployed in both consumer and industrial usage scenarios for their ubiquitous coverage, mobility support, and carrier-grade services. The resiliency of such systems is critical to sustaining billions of smartphones and IoT devices. We have identified three roadblocks hindering 5G resiliency: wireless outages, 5G software stack failures, and malicious attacks. To ensure high resiliency and build a highly available and secure 5G service, we have challenged three commonly held perceptions: 1) current failure handling at the infrastructure, modem, OS, and applications is sufficient for high resiliency; 2) providing 100\% wireless coverage ensures the best availability; 3) existing 5G security schemes, protected by SIM keys, can shield devices from attacks. Our study thus invalidates all three. The fundamental root cause stems from the complex, software-hardware interactions between devices, infrastructure, and protocol stacks. This complexity affects data transmission, exacerbates failure diagnosis and handling, and exposes new vulnerabilities to attackers.

This dissertation introduces a novel SIM/eSIM-based solution to enhance resiliency in 5G and 5G IoT systems. By utilizing the SIM card and eSIM chip as an independent miniature system, we provide plug-and-play, highly resilient 5G services without modifications to device firmware, operating systems, or base stations. We prototype and evaluate our proposals using commodity 5G devices with three major US carriers. Our solution addresses all three roadblocks of high resiliency: 1) we develop a novel SIM-based 5G failure diagnosis and handling system, resulting in 0.6x-792x disruption reduction under 5G software failures. 2) we enable rapid inter-carrier switching, reducing unavailability under outages by 28x, while only increasing power consumption by 4.7%. 3) we identify new 5G/4G attacks, such as traffic eavesdropping, man-in-the-middle attacks, and impersonation. To mitigate these threats, we offer both authentication and fine-grained access control for SIM/eSIM, ensuring security with minimal authentication latency (5.5%) and energy overhead (3.2%). By enabling intelligence on the device side with SIM/eSIM, we enhance system resiliency and address the aforementioned roadblocks.

Rethinking the "smart core, dumb terminal" design philosophy in 5G systems allows us to tap into the potential of end-user devices for improved resiliency. This traditional design principle emphasizes the importance of intelligent infrastructures while relegating end-user devices to relatively passive roles, thus neglecting their capacity to respond autonomously during service disruptions. This dissertation demonstrates that, augmenting device-side intelligence with SIMs offers a novel perspective on enhancing resiliency. SIMs enable simple, yet effective device-side handling for improved resilience by incorporating lightweight operations such as multi-tier resets and rapid multi-carrier switching. These operations allow SIMs to accelerate recovery from protocol failures, enhance connectivity outage handling, and strengthen security against vulnerabilities. By harnessing the potential of device-side intelligence, SIMs actively contribute to the resilience of both 5G and 5G IoT systems, paving the way for more robust and reliable communication networks.