量子計算下,銀行與區塊鏈面臨後量子密碼遷移難題
The Quantum Bill Just Came Due for Banks and Blockchains
IOSCO 於 11 月 19 日發布 63 頁報告,指出量子計算是金融市場一項重要的中長期網絡安全挑戰,遷移將需多年,區塊鏈尤其難以推進。Bitcoin、Ethereum 和 Solana 現用簽名為 64 bytes,ML-DSA-44 簽名約 2,420 bytes;IOSCO 引述的 Solana 測試網實驗中,改用抗量子簽名令吞吐量下降約 90%。
On November 19, the International Organization of Securities Commissions published a 63-page report with a message that was hard to miss: the financial world is not ready for quantum computers, and the upgrade it needs will be slower, more expensive, and more painful than most people in crypto want to admit.
IOSCO is not a crypto blog or a quantum startup trying to sell you something. It is the global forum of securities regulators. When it says quantum computing is a "significant medium to long-term cybersecurity challenge" for financial markets, that is the closest thing to an official warning the industry gets.
Here is the part that stuck with me. The report does not say a quantum computer exists that can break today's encryption. It says none does. What it says is that the migration will take years, that most firms have barely started, and that blockchains — the systems that pride themselves on being hard to change — will have the hardest time changing of all.
The numbers that make this concrete
A signature on Bitcoin, Ethereum, or Solana today is 64 bytes. That is the size of an elliptic-curve signature, the thing that proves you own your coins and authorizes every transaction you make.
The post-quantum replacement, ML-DSA-44, produces signatures of about 2,420 bytes. Thirty-five to forty times larger.
That is not a rounding error. On a chain, every byte is replicated to every node, stored forever, and paid for in fees. Multiply every signature on the network by 35 and the whole economics of the chain shifts. IOSCO cited a Solana testnet experiment from April where swapping in quantum-resistant signatures cut throughput by about 90 percent. A chain built on speed, running at a tenth of its speed.
I have felt this trade-off firsthand, on a much smaller scale. My own messenger uses ML-DSA-65 signatures for identity keys, and the first time I saw the key and signature sizes next to the ECDH keys they replaced, the design stopped being abstract. Post-quantum cryptography is not a drop-in replacement. It changes what your protocol can afford to do, how often you can sign, and how much data you push around. Multiply that by a global financial network and you start to see why IOSCO is worried.
Why blockchains are the hardest case
A bank can, in principle, order its vendors to swap out an algorithm. It is expensive and slow, but there is someone in charge who can sign the order. A public blockchain has no one in charge. Cryptography is baked into the core of the protocol, and changing it means coordinating thousands of independent node operators, wallet developers, exchanges, and users — many of whom have no reason to move until they are forced to.
IOSCO put it plainly: migration for blockchains may be "slow and fragmented." That is the polite version. The impolite version is that some fraction of users will never migrate, and their coins will sit in addresses whose public keys are exposed, waiting for the day a capable machine shows up. Bitcoin never mentioned the report by name — IOSCO's examples were Ethereum, Solana, and Tron — but the logic covers every chain that signs with elliptic curves. Which is all of them.
There is also a disagreement inside the crypto world about what to do with those unmigrated coins. Some argue they should be frozen or burned before a quantum computer can steal them. Others argue nobody has the right to stop a holder from spending their own coins. That argument has not been settled, and it will not be settled quietly.
Solana is actually trying
If there is a bright spot in this story, it is Solana. While most chains are still at the "we should think about this" stage, Solana's developers have a real plan with three moving parts: new signatures for users, a redesigned consensus system for validators, and a way to move existing wallets over.
The signature they picked is Falcon, not ML-DSA. The reason is size. Falcon-512 signatures are 666 bytes — still ten times larger than today's 64, but far smaller than ML-DSA's 2,420. Both of Solana's main client teams, Anza and Jump Crypto's Firedancer, studied the problem separately and landed on the same answer. The Solana Foundation said in April that network performance is "not expected to see a meaningful impact" once migration is activated. That claim will need proof, but at least someone is doing the work.
Consensus is the harder problem. In October, Solana switched on Alpenglow, a new consensus system targeting finality of around 150 milliseconds, down from roughly 12.8 seconds. Alpenglow works by bundling many validator votes into compact aggregate signatures. There is no post-quantum aggregate signature that does the same job as efficiently. So researchers at Anza published Quantumglow in July — a redesign that removes signatures wherever possible, uses a hash-based scheme called Ax, and only sends a large signature alongside small amounts of data. It is research, not a shipping feature. But it is serious research, and it exists because someone admitted the naive swap would not work.
For existing wallets, the Solana Foundation's April roadmap describes a migration path, and Anza has built a prototype that proves knowledge of the seed behind an Ed25519 account and binds it to a new post-quantum key. There is also the Winternitz Vault, an opt-in program from January 2025 that guards SOL with hash-based one-time signatures. None of this is live on mainnet yet. The point is that the groundwork is further along than the price chart suggests — SOL was trading near $69 this week, down sharply, while this work quietly continued.
The timeline nobody can agree on
IOSCO cited conservative estimates that a cryptographically relevant quantum computer could emerge between 2030 and 2035. Several industry participants it consulted said firms should prepare for the possibility of one arriving within five years, potentially sped up by AI.
Both numbers might be wrong. The honest position is that nobody knows, and the range itself is the problem. If the machine arrives in 2035, starting now looks cautious. If it arrives in 2030, starting now looks late. And "harvest now, decrypt later" means the clock is not really about when the machine arrives — it is about the data being collected today. Loan documents, merger records, private messages: anything intercepted now can be decrypted later. A German central banker illustrated a 14-year exposure window this same week. That is the number that should keep people up at night, not the qubit count.
Meanwhile, the UK's three financial regulators — the Bank of England, the FCA, and the PRA — issued joint guidance on the same day as the IOSCO report, telling firms to map their vulnerable systems, draw up replacement plans, and check their suppliers. The HKMA did the same for Hong Kong banks two weeks earlier, after scoring the sector 2.3 out of 10 on its own readiness index. The regulators are no longer asking whether to prepare. They are asking why you have not started.
What "quantum washing" tells you about where we are
My favorite detail in the IOSCO report is a warning I had not seen from a regulator before: watch out for "quantum washing" — companies overstating what their quantum products can do.
That warning only exists because the market is already full of it. Every vendor with a lattice in its logo is selling readiness. The report is, among other things, a reminder to check the claims. A library that implements ML-KEM does not make your system quantum-safe. A press release about a pilot does not make your keys safe. The gap between "we use post-quantum algorithms somewhere" and "our system survives a cryptographically relevant quantum computer" is the whole job, and most of the industry has not done it.
What I would actually do
If you build software that handles money, messages, or anything that must stay private for more than a few years, the report leaves you with a short list:
- Inventory your cryptography. You cannot migrate what you have not mapped. Find every place your system uses RSA or elliptic curves — TLS, signatures, key exchange, certificates, the lot.
- Start with the long-lived stuff. Data and signatures that must stay valid for a decade are the priority. Short-lived session keys can wait; a ten-year loan document cannot.
- Build crypto-agility in now. The algorithm you pick today might not be the one that survives. Design so you can swap it without rewriting the system.
- Do not wait for a deadline. There is no deadline. That is the trap. The migration takes years, and the only date that matters is the one you cannot see.
None of this is exciting work. It is inventory spreadsheets and migration plans and arguing with vendors. But the alternative is waking up one morning to a machine that can read everything you ever encrypted, and discovering that your upgrade path is a hard fork nobody agreed to.
IOSCO will keep quantum as a priority in 2027. The question is whether the rest of us will still be treating it as a 2035 problem by then.
Sources: IOSCO, "Implications of Quantum Computing on Financial Markets" (FR/07/26), 19 November 2026; Bank of England / PRA / FCA joint guidance on quantum computing risk, 19 November 2026; Solana Foundation quantum-readiness roadmap, April 2026; Anza / Firedancer Falcon migration research.
來源:DEV Community · dev.to