[Paper Review] Secrecy and Robustness for Active Attack in Secure Network Coding and its Application to Network Quantum Key Distribution

Masahito Hayashi, Masaki Owari+2 more|arXiv (Cornell University)|Mar 2, 2017

Cooperative Communication and Network Coding39 references3 citations

TL;DR

This paper establishes that linear network coding offers no advantage to an active eavesdropper who injects noise and monitors transmissions, proving that such attacks cannot improve information leakage beyond passive eavesdropping. It further demonstrates the existence of secure network coding protocols with asymptotic rates up to $ m_0 - m_1 - m_2 $ under secrecy and robustness constraints, and applies these results to enable secure long-distance communication via networked quantum key distribution using trusted intermediate nodes.

ABSTRACT

In network coding, we discuss the effect of sequential error injection on information leakage. We show that there is no improvement when the operations in the network are linear operations. However, when the operations in the network contains non-linear operations, we find a counterexample to improve Eve's obtained information. Furthermore, we discuss the asymptotic rate in a linear network under the secrecy and robustness conditions as well as under the secrecy condition alone. Finally, we apply our results to network quantum key distribution, which clarifies the type of network that enables us to realize secure long distance communication via short distance quantum key distribution.

Motivation & Objective

• To analyze the impact of sequential active attacks—simultaneous eavesdropping and noise injection—on information leakage in network coding.
• To determine whether non-linear operations in the network can improve an eavesdropper’s information gain compared to passive eavesdropping.
• To establish the asymptotic achievable rates for secure and robust network coding under active attack models.
• To apply the theoretical findings to network quantum key distribution (QKD), enabling long-distance secure communication using short-distance QKD links.

Proposed method

• Uses linear algebra and finite field theory to model network coding operations and analyze kernel and image relationships in the presence of noise and eavesdropping.
• Applies the universal hashing lemma and probability bounds over large finite fields $ \mathbb{F}_{q'} $ to show that the probability of successful eavesdropping drops exponentially with field size.
• Introduces a condition (F1’) involving kernel and image intersections to ensure both secrecy and robustness, and proves its sufficiency via invertible transformations.
• Employs a constructive approach to show the existence of codes with rate $ m_0 - m_1 - m_2 $ under secrecy and robustness, and $ m_0 - m_2 $ under secrecy alone.
• Analyzes a one-hop relay network model to compute mutual information $ I(M;Y_1,Y_3) $ and $ \ell_1 $-distance $ d_1(M|Y_1,Y_3) $, demonstrating $ I = 1/2 $ and $ d_1 = 1/2 $ for specific channel outputs.
• Uses symmetry and substitution (e.g., replacing $ M \to M+1 $) to extend results to other channel pairs, ensuring consistency across network configurations.

Experimental results

Research questions

• RQ1 Can an active eavesdropper who both injects noise and monitors transmissions gain more information than a passive eavesdropper in a linear network coding system?
• RQ2 Does the inclusion of non-linear operations in network coding enable an active adversary to improve their information gain beyond passive eavesdropping?
• RQ3 What is the asymptotic achievable rate of secure and robust network coding when an active adversary injects noise at rate $ m_1 $ and leaks information at rate $ m_2 $?
• RQ4 How can the theoretical results on secure network coding be applied to enable long-distance quantum key distribution using short-distance QKD links?

Key findings

• In linear networks, no active attack strategy can improve the eavesdropper’s information beyond that of passive eavesdropping, due to the invariance of linear operations under noise injection.
• For non-linear networks, a counterexample is constructed showing that active attacks can indeed increase the information leakage to the adversary.
• Under secrecy and robustness constraints, secure network coding protocols with asymptotic rate $ m_0 - m_1 - m_2 $ exist when the field size $ q' $ is sufficiently large.
• When robustness is not required and only secrecy is enforced, the achievable rate improves to $ m_0 - m_2 $, which is useful for QKD-based networks with public channel feedback.
• The probability of satisfying the key condition (F1’) for secrecy and robustness is $ 1 - O(q'^{-m_0 + m_7 - 1}) $, which approaches 1 as $ q' \to \infty $, ensuring high reliability.
• The method achieves a smaller required field dimension $ m = m_0 + 1 $ compared to prior work’s $ (m_0 - m_1)m_0 + 1 $, indicating a more efficient construction without relying on list decoding.