3D Integration: Another Dimension Toward Hardware Security

Johann Knechtel, Satwik Patnaik, Ozgur Sinanoglu

TL;DR

This paper explores how 3D integration, a technique for stacking and interconnecting multiple chips or active layers, can significantly enhance hardware security. It reviews existing security threats and schemes, arguing that 3D integration offers unique advantages for both design-time (e.g., IP piracy) and runtime (e.g., side-channel attacks) security, while also introducing new considerations.

What Problem Does This Paper Address?

The increasing complexity and globalized nature of integrated circuit (IC) design and manufacturing have created significant hardware security vulnerabilities. These threats span from design and manufacturing stages (like intellectual property (IP) piracy, hardware Trojans, and reverse engineering) to runtime operations (such as unauthorized data access, invasive probing, and side-channel attacks). Traditional security measures, often developed for 2D ICs, struggle to keep pace with these evolving threats. The paper addresses the need for novel approaches to bolster hardware security, specifically investigating how the emerging field of 3D integration can be leveraged to create more resilient and secure ICs.

Methods

The paper provides a comprehensive review of existing hardware security threats and countermeasures, categorizing them into design/manufacturing time and runtime security. It then analyzes how various forms of 3D integration—Through-Silicon Via (TSV) 3D ICs, Face-to-Face (F2F) 3D ICs, Monolithic 3D (M3D) ICs, and 2.5D ICs—can be applied to enhance these security schemes. Key proposed concepts include:

  • Secure 2.5D System-Level Integration: Utilizing an active interposer to monitor and control communication between untrusted commodity chips, acting as a security enforcement backbone.
  • "All Around" Shielding for 3D ICs: Proposing a fully enveloping 3D shield structure, built using dedicated metal layers and TSVs, to protect against invasive and non-invasive attacks.
  • Leveraging Physical Separation: Emphasizing how the inherent physical separation of components in 3D integration can be exploited to isolate trusted and untrusted elements, making attacks more difficult.

The paper also discusses the integration of existing security techniques like logic locking (LL), layout camouflaging (LC), and split manufacturing (SM) within the 3D integration paradigm.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • 3D Integration as a Security Enabler: The core insight is that 3D integration is not just about performance or density; it offers a fundamentally new dimension for enhancing hardware security by enabling novel architectural approaches.
  • Physical Separation is Key: The ability to physically separate components (chips, layers, or functional blocks) in 3D integration is a powerful tool for security. This allows for the isolation of untrusted elements from trusted security modules.
  • Runtime Security Benefits: 3D integration is particularly promising for runtime security, enabling concepts like comprehensive "all-around" shielding that are difficult or impossible to achieve in 2D.
  • New Threats and Opportunities: While 3D integration enhances security, it also introduces new attack vectors and requires careful consideration of how components are partitioned and interconnected to avoid creating new vulnerabilities.
  • 2.5D for System-Level Security: The paper highlights 2.5D integration with an active interposer as a practical approach for system-level security, allowing untrusted chips to be integrated while a trusted interposer monitors their behavior.

Key Results

The paper is primarily a survey and conceptual proposal, rather than a presentation of empirical results from new experiments. However, it synthesizes findings from prior work, indicating:

  • Improved Scalability for Security Techniques: Monolithic 3D (M3D) ICs can make security techniques like layout camouflaging more scalable and cost-effective compared to 2D implementations, with reported average savings of 25% power, 15% delay, and 43% area.
  • Enhanced Split Manufacturing: 3D integration makes split manufacturing more practical by allowing different parts of the system (e.g., FEOL and BEOL) to be manufactured independently across multiple chips, offering greater flexibility in withholding functional components from untrusted parties.
  • Novel Shielding Concepts: The paper proposes an "all around" shielding concept for 3D ICs, which, if implemented, would offer a robust defense against invasive and non-invasive attacks by creating a complete physical barrier.
  • System-Level Monitoring: The proposed 2.5D secure integration concept aims to provide a system-level monitoring capability for untrusted chips, enhancing security without necessarily modifying the untrusted chips themselves.

Limitations and Future Work

  • Complexity of 3D Integration: The paper acknowledges that the complexity of 3D integration itself can introduce new vulnerabilities. The physical implementation of security features needs careful design to avoid becoming a weakness.
  • Cost-Security Trade-offs: The paper notes that the degree of partitioning in 3D split manufacturing impacts cost and security. More splitting can increase layout costs due to more vertical interconnects but offers greater security flexibility.
  • New Attack Vectors: The authors caution that the broader landscape of suppliers and actors involved in 3D integration can open up new opportunities for attackers to embed different types of hardware Trojans.
  • Future Work: The paper suggests further research into the practical implementation and validation of the proposed 2.5D system-level integration and "all around" shielding concepts. It also calls for more investigation into the specific threats and countermeasures related to hardware Trojans in 3D ICs.

Summary updated

Abstract

We review threats and selected schemes concerning hardware security at design and manufacturing time as well as at runtime. We find that 3D integration can serve well to enhance the resilience of different hardware security schemes, but it also requires thoughtful use of the options provided by the umbrella term of 3D integration. Toward enforcing security at runtime, we envision secure 2.5D system-level integration of untrusted chips and “all around” shielding for 3D ICs.
BibTeX
@inproceedings{knechtel2019d,
  author = {Johann Knechtel and Satwik Patnaik and Ozgur Sinanoglu},
  title = {{3D Integration: Another Dimension Toward Hardware Security}},
  booktitle = {IOLTS},
  year = {2019},
  url = {https://openreview.net/forum?id=608hARiYBF},
}

Loading related content...

Figure