I Didn’t Take Harvest-Now-Decrypt-Later Seriously Until I Ran the Math on My Own Data

I spent years writing about network security without giving harvest-now-decrypt-later more than a passing mention. It sounded like a problem for cryptographers in 2035, not something I needed to think about on a Tuesday. Then I sat down and actually ran the math on a client's data retention policy, and realized the exposure window was already open.

What Harvest-Now-Decrypt-Later Actually Means

Harvest-now-decrypt-later, HNDL for short, describes an attacker copying encrypted network traffic today and sitting on it until a future computer, quantum or otherwise, can break the key exchange that protected it. The attacker isn't trying to crack your encryption this afternoon. They're betting on a capability that might not exist for another five or ten years, which is exactly why this feels like tomorrow's problem when it's actually today's.

The part that changed my thinking is the timing gap. The theft happens now, quietly, while the payoff arrives whenever the attacker's tools catch up. Security researchers use a rough test for whether a given data stream is already exposed: add up how long that data needs to stay confidential, add the time your organization will need to migrate to safer encryption, and compare that total against how many years remain before quantum computers can plausibly break today's algorithms. For anything with a decade-plus shelf life, government records, financial data, long-term intellectual property, that math often comes out upside down already.

Why Nobody Notices When It Happens

What makes HNDL different from a normal breach is the silence. There's no failed login, no strange access pattern, no log entry that trips an alert. An attacker taps a fiber optic line, physically, and copies the signal without touching anything on your network that would notify a security team. Intelligence and telecom researchers have documented years of exactly this kind of interception, from trunk-line taps to submarine cable incidents to street-cabinet and junction-box access on carrier-grade traffic, and encryption alone doesn't remove that exposure. A tapped signal that's merely encrypted, rather than made physically unreadable, is still a complete, storable copy of your traffic sitting in someone else's archive, and there's no retroactive fix once it's been recorded.

That last part is what stuck with me. Most security thinking assumes you can respond after the fact, patch the hole, rotate the credentials, notify the people affected. HNDL breaks that assumption. By the time anyone could plausibly detect that a signal was harvested, the copy has already been sitting in storage for however many years it takes for the decryption side of the equation to catch up.

Do you know I Used to Think Encrypted Fiber Was Enough. Then I Learned What Harvest Now, Decrypt Later Really Means? It's a good look at how someone who assumed encryption was enough had to rethink that after learning what a tapped fiber line can still hand an attacker.

The Part Post-Quantum Cryptography Doesn't Fix

Once I understood the recording problem, I went looking for the fix, and most of what I found was post-quantum cryptography. PQC replaces the vulnerable RSA or elliptic-curve key exchange with algorithms like ML-KEM that are believed to resist quantum attacks. That's real, necessary work, and organizations should be doing it now for every TLS, VPN, and IPsec deployment they run.

But PQC only protects keys generated after you've deployed it. It does nothing about a signal that's already been recorded off the wire, and it does nothing to stop a signal from being recorded going forward, because the payload itself, even when AES-encrypted, is still a coherent, capturable waveform on the fiber. The weak link was never AES. It's the key exchange, and swapping which algorithm produces that key exchange doesn't change the fact that the traffic can still be tapped and archived.

Wondering what it actually looks like when fiber data becomes physically unrecordable? How CyberRidge Makes Fiber Data Unrecordable, Even If It's Tapped walks through it.

Where CyberRidge Fits

This is the gap CyberRidge built its Carmel platform to close. Instead of trying to make the encryption unbreakable forever, Carmel goes after the recording step directly. It spreads the optical signal across a wide spectral band, applies a constantly changing optical key generated inside the light stream itself, and buries the result under a layer of controlled optical noise. What an attacker intercepts isn't a structured, storable signal anymore; it's noise with no usable data in it, regardless of what computing power shows up in ten years.

Carmel installs as a drop-in replacement for a standard transmission line card, which is the detail that made me take it seriously. It's not a network redesign. CyberRidge reports deployments completing in weeks with no added latency, and the platform is built to complement PQC and other higher-layer protocols rather than replace them. The two problems, breakable keys and recordable signals, need different fixes, and Carmel is aimed squarely at the second one.

What I'd Actually Do About It

If any of your data needs to stay confidential for five years or more, run the shelf-life-plus-migration-time math yourself before assuming this doesn't apply to you. Start post-quantum migration now regardless, since that clock doesn't get shorter by waiting, and it's the required baseline no matter what else you add on top of it. And for the specific links carrying your longest-lived sensitive data, look seriously at physical-layer protection, because it's the only piece of this stack that stops the recording itself rather than just racing to outrun it.

None of this needs to be an overnight overhaul. Start with an honest inventory of which links actually carry the long-shelf-life data, prioritize those, and treat the rest as normal PQC migration work on its usual timeline. The mistake I nearly made was treating HNDL as an all-or-nothing problem instead of one you can triage like anything else.

FAQs

Q: What's the difference between HNDL and a normal data breach?

A: A normal breach usually leaves some trace, a login anomaly, a log entry, an alert. HNDL is a passive fiber tap that copies traffic without touching your network defenses at all, so there's typically no way to know it happened until the encryption protecting that traffic is eventually broken.

Q: Does post-quantum cryptography fully solve harvest-now-decrypt-later?

A: No. PQC protects the key exchange for data transmitted after it's deployed, but it doesn't do anything about traffic that's already been recorded, and it doesn't stop a signal from being tapped and stored going forward. The payload itself remains a recordable waveform either way.

Q: How do I know if my organization's data is already exposed?

A: Compare how long your data needs to stay confidential plus your realistic migration timeline against conservative estimates for when quantum computers could break today's key exchange algorithms. For data with a shelf life past 2030 or so, many teams find the exposure window is already open.

Q: What does CyberRidge's Carmel platform actually do?

A: Carmel is a physical-layer defense that spreads an optical signal across a wide spectral band, embeds a constantly changing optical key inside the transmission, and buries the result under controlled noise, so an intercepted signal has nothing usable in it to record.

Q: Does Carmel replace the need for post-quantum cryptography?

A: No. CyberRidge positions Carmel as complementary to PQC and other higher-layer protocols, not a substitute. PQC protects future key exchanges; Carmel prevents the underlying signal from being a usable recording in the first place.

Q: How long does it take to deploy Carmel?

A: CyberRidge reports deployment in weeks, since it installs as a drop-in replacement for an existing transmission line card rather than requiring a redesign of the surrounding network. Actual timelines will vary with network complexity.