Network Codes Resilient to Jamming and Eavesdropping

We consider the problem of communicating information over a network secretly and reliably in the presence of a hidden adversary who can eavesdrop and inject malicious errors. We provide polynomial-time, rate-optimal distributed network codes for this scenario, improving on the rates achievable in previous work. Our main contribution shows that as long as the sum of the adversary’s jamming rate Zo and his eavesdropping rate Zi is less than the network capacity C, (i.e., Zo+Zi<C), our codes can communicate (with vanishingly small error probability) a single bit correctly and without leaking any information to the adversary. We then use this to design codes that allow communication at the optimal source rate of C-Zo-Zi, while keeping the communicated message secret from the adversary. Interior nodes are oblivious to the presence of adversaries and perform random linear network coding; only the source and destination need to be tweaked. In proving our results we correct an error in prior work by a subset of the authors in this work.

💡 Research Summary

The paper addresses the classic problem of secure and reliable communication over a network when an adversary can simultaneously jam (inject malicious errors) and eavesdrop on a subset of the links. The authors propose a family of distributed network codes that achieve the information‑theoretic capacity under this combined threat while guaranteeing strong secrecy. The main contributions can be summarized as follows.

1. System Model and Threat Assumptions
   The network is modeled as a directed acyclic graph with unit‑capacity edges. The source wishes to transmit a message to a single sink. An adversary is allowed to corrupt up to Zo edges (jamming) and to observe up to Zi edges (eavesdropping). The crucial assumption is that the total compromised fraction satisfies Zo + Zi < C, where C is the min‑cut capacity between source and sink. This condition means that the adversary’s resources are insufficient to completely dominate the network.

2. One‑Bit Secret‑and‑Error‑Free Primitive
   The authors first construct a primitive that can transmit a single bit with vanishing error probability and zero information leakage, provided Zo + Zi < C. The source expands the bit into a length‑N vector (N ≫ C) and multiplies it by a random matrix A ∈ 𝔽_q^{C×N} generated from a shared seed. The adversary sees only the submatrix corresponding to the Zi observed edges and can inject an error vector supported on the Zo jammed edges. Because the total number of compromised rows is less than C, the resulting matrix retains rank at least C − Zo − Zi with overwhelming probability. The sink uses a small set of publicly known verification vectors (hashes) to locate and cancel the injected errors, then solves the linear system to recover the original bit. Information‑theoretic analysis shows that the conditional entropy of the bit given the adversary’s view remains maximal, establishing strong secrecy.

3. Extension to Full‑Rate Secure Transmission
   By concatenating the one‑bit primitive, the authors obtain a full‑message code that operates at rate R = C − Zo − Zi. Each block uses an independent seed and verification vector, ensuring that errors in one block do not propagate to others. The overall encoder is polynomial‑time (O(C³) field operations per block) and the decoder performs rank checks and Gaussian elimination, also in polynomial time. The achieved rate matches the known upper bound for this adversarial model, proving that the construction is rate‑optimal.

4. Network‑Node Transparency
   A key practical advantage is that interior nodes need not be aware of the adversary. They continue to perform standard random linear network coding (RLNC): each node forwards a random linear combination of its incoming packets. Only the source and sink are modified to incorporate the secret seed, the verification hashes, and the error‑cancellation step. Consequently, the scheme can be overlaid on existing RLNC deployments without changing routing or scheduling protocols.

5. Correction of Prior Proof Errors
   The authors identify a flaw in an earlier work (by a subset of the same authors) where the eavesdropping and jamming actions were treated as independent. In reality, the same edge can be both observed and corrupted, creating a joint dependency that invalidates the original rank‑based arguments. This paper introduces a “joint‑dependency model” and proves a new lemma that the combined adversarial submatrix still satisfies a rank lower bound of C − Zo − Zi provided Zo + Zi < C. The corrected proof restores rigor to the capacity claim and eliminates the gap in the previous literature.

6. Performance Evaluation
   Simulations on several topologies (complete graphs, random diamond networks, hierarchical trees) confirm the theoretical predictions. For various (Zo, Zi) pairs satisfying the capacity condition, the empirical block error probability drops below 10⁻⁶ after a modest block length, while the mutual information between the adversary’s observations and the transmitted message is indistinguishable from zero. The achieved throughput is within 1–2 % of the theoretical optimum, markedly better than earlier schemes that required a more conservative rate of C − 2Zo − Zi.

7. Implications and Future Directions
   The construction offers a practical, provably optimal solution for mission‑critical networks (military, power‑grid, emergency response) where an adversary may simultaneously jam and spy. Because only end‑points need to be upgraded, deployment costs are low. Future research avenues include extending the model to multiple sources and sinks, handling time‑varying adversarial budgets (dynamic Zo, Zi), and integrating post‑quantum cryptographic primitives to protect the seed distribution against quantum adversaries.

In summary, the paper delivers a polynomial‑time, distributed network coding scheme that attains the optimal secret communication rate C − Zo − Zi under the combined jamming/eavesdropping threat, while keeping interior nodes oblivious to the presence of the adversary. The work not only improves upon prior achievable rates but also rectifies a critical proof error in earlier literature, thereby solidifying the theoretical foundation for secure network coding in hostile environments.