TL;DR
This paper argues for integrating security as a first-class citizen in the early design phases of embedded systems through security-aware design-space exploration (DSE). It proposes a multifaceted, scoring-based methodology to quantify the "degree of secureness" of design instances, enabling multi-objective optimization that balances security with performance, power, and cost.
What Problem Does This Paper Address?
Modern embedded systems are increasingly complex and interconnected, making them attractive targets for attackers. While traditional design objectives like performance and power consumption are considered early in the design process through Design-Space Exploration (DSE), security is often treated as an afterthought. This reactive approach leads to suboptimal designs because security measures implemented late in the design cycle can negatively impact established trade-offs with other crucial objectives. The core problem is the lack of a systematic way to quantify and integrate system security into the early, multi-objective optimization process of DSE.
Methods
The paper proposes a "security-aware, system-level DSE" approach. The core of this approach is a multifaceted, scoring-based methodology for secureness quantification. This methodology involves:
- Identifying Application Requirements: Specifying extra-functional requirements, including those related to security (Confidentiality, Integrity, Availability - CIA triad).
- Determining Domain-Specific Attacks: Identifying relevant attack types (e.g., side-channel, DoS, software-based, cryptanalysis) based on the system's architecture and security requirements.
- Mapping Attacks to Components: Determining which components are susceptible to which attacks.
- Identifying Protection Mechanisms: Selecting relevant protection mechanisms for each attack type.
- Quantifying Secureness: Developing a scoring mechanism for individual components and then aggregating these scores for the entire design instance. This scoring considers:
- Protection Level: The extent to which a protection mechanism is implemented for a given attack.
- Utilization: The usage level of a component, with the rationale that lower utilization might imply lower susceptibility to certain attacks.
- (Future Work/Implied) Spatial Isolation: Reducing resource sharing between applications or tasks to enhance security.
The paper outlines a potential formula for component security scoring that incorporates protection level and inverse utilization. The overall security score for a design instance is then the accumulation of component scores.
Datasets
Not discussed in the paper.
Key Insights and Ideas
- Security as a Design Objective: The paper strongly advocates for treating security not as an add-on feature but as a fundamental design objective, on par with performance and power consumption, from the earliest stages of development.
- The Cost of Late Security Integration: Delaying security considerations until later design phases can lead to suboptimal system designs due to unforeseen trade-offs with other critical objectives.
- Quantifying "Secureness": The central challenge is the lack of a universally accepted metric for "degree of secureness." The proposed scoring methodology aims to bridge this gap by providing a quantifiable measure.
- Multifaceted Scoring: A single metric is insufficient. The proposed approach considers multiple factors (protection mechanisms, utilization) to provide a more holistic view of security.
- Enabling Trade-off Analysis: By quantifying security, designers can explicitly study and optimize the trade-offs between security and other extra-functional properties, leading to better-informed design decisions.
Key Results
This paper is a position paper and does not present an actual implementation or experimental results. Therefore, there are no quantitative results to summarize. The key "result" is the argument and proposal for a new methodology that, if implemented, would enable:
- Optimization of security aspects during the earliest design phases.
- Study of trade-offs between security and other design objectives (performance, power, cost).
- A substantial competitive advantage for companies in the embedded systems industry.
Limitations and Future Work
The authors explicitly state that this is a position paper and the proposed methodology has not yet been implemented. They acknowledge that:
- The scoring example provided is a rough sketch and not a complete solution.
- Spatial isolation is mentioned as a factor but not fully integrated into the scoring formula in the presented example.
- Further research is needed to implement and validate the proposed scoring methodology.
- The paper does not delve into the specifics of how the domain-specific attacks and protection mechanisms would be exhaustively identified and cataloged for all possible embedded systems.
The authors hope this paper will trigger more widespread research into incorporating security as a first-class citizen in early design-space exploration.
Summary updated