A lot of people focus on "harvest now, decrypt later" attacks in which encrypted data is gathered now and eventually decrypted once a sufficiently powerful quantum computer is available. However, there's nothing new about this attack in that it has existed for decades. I've also found that this explanation rarely resonates with engineers who believe a powerful quantum computer is decades away based on current progress. While I'd argue that it doesn't matter how any of us feel as NIST and other government bodies have mandated a transition to post-quantum cryptography [1], that's the subject of a different post.

Thus, the more interesting question that I'll seek to answer is: why should we care about quantum computing now and why can't we just migrate once we're closer to having a powerful quantum computer?

Technical people and friends alike ask me this so frequently that it's worth answering with two brief history lessons that combine to make my point.

WWII, Enigma, and Alan Turing

Cryptography has long played an instrumental role in the military because different units inevitably need to communicate with one another across a distance. Some of the first known "ciphers" were used by the Romans to share military positions (i.e. Caesar cipher). The assumption was that any message could be intercepted by the enemy so communication should only be "readable" by the right people who knew the "key" to decipher it.

During WWII, the Nazi military used the Enigma machine to encrypt messages that they knew were being intercepted by the Allied forces.

In an extremely classified effort spanning years at Bletchley Park, Alan Turing (working alongside many others) "broke" the Enigma cipher, enabling the Allied forces to read intercepted Nazi communication about their war plans [2]. However, the Allies had to disguise their behavior so that the Nazis would not realize this had happened and change to a new cipher [3].

The information gleaned from these intercepted communications was so valuable that both Churchill and Eisenhower credited the breaking of Enigma with helping the Allies to win the war. However, it was only revealed in 1974 (almost 3 decades after the war ended!) that Enigma had been "broken." There is much speculation as to why this is but that's a topic of a different post. Moreover, some of Alan Turing's papers detailing the mathematical techniques he used to break Enigma were not released to the public until 2012 as they were presumably still valuable.

Scientific progress is a step function

Scientific progress often behaves like a step function, especially when it comes to theoretical computer science, math, and physics. There are periods where a lot of small improvements happen over a relatively short timeframe but there are also long quiet stretches punctuated by giant abrupt leaps.

We can see some of this in the last few years with the progression of machine learning, such as OpenAI solving the Navier-Stokes problem [4] and Anthropic discovering a major attack on what was thought to be a very secure post-quantum digital signature scheme in consideration for standardization.

With the advent of AI, it is very possible (if not extremely likely) that scientific progress accelerates and R&D timelines compress.

Furthermore, there is not just "one front" in which progress can happen for quantum computing. When speaking with engineers, I noticed most are only aware of the hardware side of things (e.g. asking how we'll reach the low temperatures needed for superconducting processors). There's a general lack of awareness that progress could happen on two additional fronts, namely quantum algorithms or quantum error correction.

On the algorithmic front, progress might look like reducing the number of logical qubits or gates needed in the algorithm so that it's easier to implement on the physical quantum computer (since needing fewer logical qubits translates to needing fewer physical qubits in the instantiation). On the even lesser known quantum error correction front, this might look like being able to correct for larger amounts of error or having more efficient codes with less redundancy (which would also translate into needing fewer physical qubits in the quantum computer).

Progress may be extremely challenging on the hardware front (i.e. building the physical quantum computer). However, it would be unsurprising, if not even expected, for huge progress to be made on the more theory-oriented fronts of quantum algorithms and error correction, especially with the aid of AI.

So why am I telling you all of this?

A powerful quantum computer may be even more valuable than breaking Enigma, especially if there are enough companies still using quantum-vulnerable cryptography. In addition to reading communication between various parties, an attacker can take a more active role by forging digital signatures to extract information from competitors, mislead service providers, or steal seemingly endless amounts of cryptocurrency. With the gap between theory and implementation narrowing, governments want everyone to transition to post-quantum cryptography sooner rather than later.

At the point a nation state or an equally rich corporation has a working practically interesting quantum computer (sometimes referred to as Q-day [5]), they may not announce that to anyone. If I were running such an effort, I would certainly not advertise what a huge advantage I had over everyone else. In fact, I would operate exactly as the Allies did in Bletchley Park, reading whatever I wanted and disguising my behavior to hide that fact I had such a superpower.

That takes us back to the title of this post: what is the real quantum threat? The real quantum threat is that we're a lot closer to Q-day than the rest of us know, that the actual research progress is not being shared externally, and that when Q-day does finally come we'll be none the wiser.

Notes

[1] NIST (the federal agency that determines cryptographic standards) has disallowed quantum-vulnerable cryptography after 2035. Furthermore, the transition to post-quantum cryptography is expected to take years for most companies to complete as it requires taking an inventory of existing software (i.e. producing a cryptographic bill of materials) and determining an appropriate migration path.

[2] I am vastly simplifying the story here. Some important pieces omitted include that a number of Polish individuals helped lay the foundation for the breaking of the Enigma cipher as well as that different branches of the Nazi military used different Enigmas.

[3] If you want a dramatic (but not entirely factual) retelling of this story, I'd recommend watching The Imitation Game.

[4] There is a lot of controversy around OpenAI and their solution to the Navier-Stokes problem (e.g. whether OpenAI made use of another mathematician's work, what "version" of the problem did OpenAI solve and whether it is useful).

[5] More specifically, Q-day is the day on which quantum computers are powerful enough to "break" modern-day public key cryptography that is not post-quantum, such as ECDSA and RSA.

‍