Secure Upgrade of Hardware Security Modules in Bank Networks

Riccardo Focardi, Flaminia L. LuccioARSPA-WITS 2010 · DOI

TL;DR

This paper tackles the challenge of securely upgrading critical hardware security modules (HSMs) in bank networks, which are vulnerable to insider attacks. They propose a novel incremental upgrade strategy that balances security improvements with network compatibility and cost-effectiveness, drawing an interesting parallel to graph decontamination problems.

What Problem Does This Paper Address?

Hardware Security Modules (HSMs) are vital components in bank networks, responsible for securely processing and verifying Personal Identification Numbers (PINs) from ATMs. However, these HSMs have been found to be vulnerable to API-level attacks, particularly by insiders who can exploit these weaknesses to recover user PINs and potentially clone cards. While fixes exist, such as modifying HSM functionality or adding integrity checks, a complete worldwide upgrade of all HSMs is prohibitively expensive and complex. The core problem is how to upgrade these critical systems incrementally, ensuring compatibility with the existing legacy infrastructure while progressively enhancing security, all within practical cost constraints.

Methods

The authors propose an incremental upgrade strategy that allows for the creation of secure subnets within the larger network, maintaining compatibility with the legacy system. This is achieved by introducing "borderline HSMs" that act as translators between upgraded and non-upgraded nodes. The core of their approach involves formalizing the HSM upgrading problem as an algorithmic problem on trees and demonstrating its equivalence to the Connected Monotone Decontamination (CMD) problem. They adapt algorithms from CMD to solve the HSM upgrading problem, focusing on minimizing two key parameters: the number of borderline HSMs (B) and the number of technicians (U) involved in the upgrade process. They develop rules and an algorithm (Algorithm UPGRADE) to determine the optimal placement of borderline HSMs and the sequence of upgrades, considering trade-offs between B and U.

Datasets

Not discussed in the paper.

Key Insights and Ideas

  • Incremental Upgrade Strategy: The paper introduces a practical approach to upgrading critical infrastructure by allowing for phased upgrades, creating secure "islands" within a larger, less secure network.
  • Borderline HSMs: The concept of borderline HSMs is crucial. These act as intermediaries, enabling communication between upgraded and legacy systems, thus preserving network functionality during the transition.
  • Equivalence to Graph Decontamination: A significant insight is the formal mapping of the HSM upgrading problem to the Connected Monotone Decontamination (CMD) problem. This allows leveraging existing, well-studied algorithms and theoretical results from graph theory.
  • Trade-off Optimization: The paper frames the problem as optimizing a trade-off between the cost of hardware (borderline HSMs) and the cost of labor (technicians), providing a framework for making informed decisions about upgrade strategies.
  • Algorithmic Adaptation: The authors successfully adapt algorithms from the CMD problem to solve the HSM upgrading problem, demonstrating the power of cross-domain problem mapping.

Key Results

The paper establishes an equivalence between the HSM upgrading problem and the Connected Monotone Decontamination (CMD) problem. They prove that for a single technician (U=1), the minimal number of borderline HSMs required is directly related to the connected search number of the network graph. They present an algorithm (Algorithm UPGRADE) that correctly and optimally solves the HSM upgrading problem for tree networks, minimizing the number of borderline HSMs. Furthermore, they introduce a modified rule (Rule 1 new) to account for multiple technicians (U > 1), showing how increasing the number of technicians can reduce the number of required borderline HSMs. An example illustrates how to estimate upgrade costs by balancing the cost of HSMs against the cost of technicians.

Limitations and Future Work

  • Network Topology: The current work focuses on tree networks. The authors acknowledge that real-world bank networks are more complex and suggest investigating other topologies as future work.
  • Borderline HSM Placement: The paper primarily considers placing borderline HSMs on nodes. They suggest exploring the placement of borderline HSMs on edges as a future direction, which might increase hardware requirements but potentially offer better security.
  • Technician Motion: The current model assumes technicians can move freely within the network. The authors propose optimizing technician motion, especially in weighted networks (representing physical distances), as a future research avenue, noting its potential complexity similar to the Traveling Salesman Problem.
  • Degree of Security: The paper focuses on functional upgrade and security against specific attacks. Quantifying and comparing the "degree of security" provided by different strategies (e.g., edge vs. node placement of HSMs) is identified as an area for future investigation.

Summary updated

BibTeX
@inproceedings{focardi2010secure,
  author = {Riccardo Focardi and Flaminia L. Luccio},
  title = {{Secure Upgrade of Hardware Security Modules in Bank Networks}},
  booktitle = {ARSPA-WITS},
  year = {2010},
  doi = {10.1007/978-3-642-16074-5\_7},
}
Figure