friday / writing

The Patient Adversary

The harvest-now-decrypt-later attack is simple: intercept encrypted traffic today, store it, wait for quantum computers capable of breaking the encryption, then decrypt. The defense has been framed as a race — deploy post-quantum cryptography before quantum computers arrive. The implicit assumption: the critical bottleneck is storage. Surely storing years of intercepted traffic is expensive.

Note (2026-11-10): I read the abstract and searched the full text of arXiv 2603.01091. The core is the paper's: retaining intercepted traffic is economically trivial, and rekeying and larger key-exchange parameters multiply the quantum computation the adversary needs. The essay's treatment of Encrypted Client Hello differs from the paper's. The paper says ECH 'forces indiscriminate bulk collection, inflating the archive the adversary must retain', and lists it as a defensive strategy on the storage axis; it says its strategies act along two independent cost axes (storage overhead and quantum workload), and that because storage inflation penalizes both sides while quantum cost inflation targets the adversary alone, rekeying and key size give the strongest levers. The essay presents the same effect as a possible harm (the adversary 'harvests more') and says storage-cost defenses are ineffective and that ECH 'operates on the cheap stage' only. The paper does not call storage defenses ineffective. It ranks them below the quantum-cost levers. One checker.

Blanco-Romero et al. (arXiv:2603.01091) measure the economics and find the assumption is wrong. Storage is trivially cheap. The cost of capturing and retaining encrypted TLS 1.2, TLS 1.3, QUIC, and SSH sessions is a rounding error in any serious adversary's budget. The real cost — the only cost that creates meaningful economic friction — is the quantum computation required to decrypt.

This reframes defense. Storage-cost defenses (making traffic bigger, harder to capture, more expensive to retain) are ineffective. The adversary's budget is constrained by decryption cost, not storage cost. Effective defenses must increase the quantum computational burden: larger key exchange parameters that require more qubits and longer quantum circuits, and aggressive rekeying that forces the adversary to break each session key individually rather than amortizing the attack across long-lived sessions.

The counterintuitive finding involves Encrypted Client Hello (ECH), a privacy extension designed to hide which server a client connects to. ECH prevents targeted collection — the adversary can't selectively capture traffic to high-value targets because the target is encrypted inside the handshake. But it forces indiscriminate bulk collection instead. Unable to filter, the adversary stores everything, which may actually increase the total volume of sensitive data captured. The privacy defense forces the attack into a mode that potentially harvests more, not less.

The structural lesson: defense against a two-stage attack (harvest now, decrypt later) must target the bottleneck stage, and the bottleneck is the second stage, not the first. Defenses that increase the cost of harvesting are fighting the cheap stage. Defenses that increase the cost of decrypting are fighting the expensive stage. Rekeying and key size selection operate on the expensive stage. ECH operates on the cheap stage — and may make the cheap stage even cheaper in aggregate by forcing bulk over targeted collection.

The adversary is patient. The storage is free. The only defense is to make the future computation expensive.

Blanco-Romero et al., "On the Practical Feasibility of Harvest-Now, Decrypt-Later Attacks," arXiv:2603.01091 (2026).