Trusted computing technology provides promising new security primitives for the development of dependable and highly secure distributed applications. Protected software components, or enclaves, are provided with strong integrity, confidentiality and key derivation mechanisms via a minimal, hardware-only trusted computing base. Yet, leveraging enclaved execution in practice remains not well-understood, especially for heterogeneous trusted computing networks. This tutorial aims to provide insights and hands-on exercises on building such interactive applications based on Intel SGX and the lightweight embedded Sancus architecture.

Trusted computing architectures such as Intel SGX, ARM TrustZone, and Sancus have been around for a number of years. By enforcing strong integrity, confidentiality, and attestation guarantees with a minimal (hardware-only) trusted computing base, these architectures aim to provide a root-of-trust for the development of dependable and highly secure systems. However, few real-world applications leverage trusted computing, and approaches to interconnect enclave applications in heterogeneous and distributed trusted computing environments remain not well-understood. Yet, this approach provides strong security benefits in domains like the Internet of Things or interconnected Industrial Control Systems.

This tutorial outlines the development of such security-sensitive networked applications. Participants will learn how to write enclaved software modules for both the embedded open-source Sancus research architecture, as well as for off-the-shelf Intel SGX platforms. We will show how to make Sancus/SGX enclaves interact securely in a distributed fashion. Our approach is distinguished by the concept of authentic execution, a notion of end-to-end security and secure I/O where the overall application's behavior depends solely on protected device drivers and data processing enclaves.

The tutorial will cover common pitfalls for enclave development and feature hands-on exercises on programming, deploying, attesting, and interacting with basic Sancus/SGX enclaves. The tutorial setup uses an automotive industry-standard CAN bus to interconnect Intel SGX machines with Sancus-enabled microcontrollers. We will provide Sancus I/O driver modules plus a CAN authentication library that allows participants to easily setup secure communication channels.

Our tutorial aims at researchers, software developers, and software architects who want to learn more about modern trusted computing architectures. Basic programming skills (in C) and familiarity with the GNU/Linux command line interface are required to follow the hands-on exercises.

Participants will be provided with remote SSH access to an SGX-capable machine, connected to Sancus-enabled microcontrollers over a CAN network. This machine will also have pre-installed the required software development tools for the two trusted computing architectures, as well as skeleton source code for the hands-on exercises.

The tutorial aims to fill about three hours with an equal split between lectures and practical exercises:

Duration	Content
45 min	Introductory lecture: Trusted Computing and remote attestation Authentic Execution on heterogeneous Trusted Computing infrastructure
45 min	Hands-on: Developing Sancus/SGX enclaves, software deployment, remote attestation and secure communication
45 min	Hands-on: Secure I/O on Sancus, authentic execution in a distributed setting
45 min	Pitfalls, advanced topics, wrap-up Pitfalls: untrusted pointers, function pointers, resource sharing, debugging Compartmentalising applications Side-channels and how to avoid them Open research questions