Foundations for Sustainable Security

Introduction

Security and efficiency are often seen as a conflict. IT already consumes 11% of electricity globally, with a steep upwards trend. Resource sharing increases efficiency but introduces information leakage vulnerabilities, such as Meltdown and Spectre. Reducing reliability margins also increases efficiency but introduces fault attacks, such as Rowhammer and Plundervolt. This reveals a fundamental problem in current systems: reliability mechanisms are not designed with adversaries in mind. Security is then patched on top of reliability mechanisms, incurring additional energy costs.

Objectives

We will overcome the conflict between security and efficiency with novel foundations to make security sustainable and use security to increase efficiency. Our goals include:

1. Researching how to measure the efficiency of security.
2. Designing principled and efficient security mechanisms.
3. Utilizing security to increase efficiency.
4. Securing microarchitectural optimizations.
5. Securing lightweight isolation.

Methodology

Our methodology is to integrate principled cryptography-grade security into all system layers to minimize and supersede inefficient reliability mechanisms. Key components of our approach include:

• Developing a framework for fine-grained energy efficiency measurements.
• Researching fine-grained replication for side-channel isolation while maintaining efficiency.
• Exploring selective resource sharing for secure variables, enclaves, and virtual machines, superseding today's inefficient and insecure techniques.

Originality

The originality of FSSec stands out in that energy efficiency has played no role in security so far. In particular, using cryptography to replace established error correction methods will be the key to our goal of using security to increase efficiency by 20% compared to current systems. We will construct secure optimizations with fine-grained isolation, increasing efficiency without adding side channels.

Team

Asst.-Prof. Daniel Gruss heads an internationally renowned security research group. FSSec will fund 6 PhD students.