Encryption Myths vs Facts: How the Basics Changed Since 2020

Archi4 min read

cryptographysecuritystandards

Contents
  1. The rules of encryption did not change. The deadlines did.
  2. Myth: Nothing important changed since 2020
  3. Myth: TLS 1.3 is still the new thing
  4. Myth: Hashing is boring and settled
  5. Myth: Quantum computing is a distant problem
  6. Myth: Longer keys mean stronger security
  7. What to check this quarter

The rules of encryption did not change. The deadlines did.#

Five years ago, a typical engineering team deployed TLS 1.2, rotated keys when someone remembered, and treated hashing as a solved problem. That description still fits many teams in 2026. The difference is that the calendar is now doing the arguing. Standards bodies have set dates for retiring old algorithms, vendors have started shipping post-quantum key exchange by default, and auditors ask for documents that did not exist in 2021.

Here are the myths that survive, checked against where things stood five years ago.

Myth: Nothing important changed since 2020#

Fact: The primitives are stable. The deployment patterns are not.

SHA-256 and AES-256 look the same on paper. What changed is how often they run. Five years ago, a web service might terminate a few thousand TLS sessions per hour. Today the same service terminates millions, many of them machine to machine, and the certificate chain has become a performance line item rather than an afterthought.

Key exchange is the clearest break. In 2021, hybrid post-quantum key exchange was a lab experiment run by a handful of large providers. By 2025 it was on by default for a large share of browser traffic. The handshake still completes in milliseconds, but the bytes on the wire are different, and old middleboxes that inspect traffic have started to choke on the larger ClientHello. If a load balancer was configured in 2019 and never touched, it is now a likely failure point.

Myth: TLS 1.3 is still the new thing#

Fact: TLS 1.3 is the floor, and the interesting work moved up the stack.

TLS 1.3 was finalized in 2018. Five years ago, adoption was climbing fast but uneven, and plenty of internal services still spoke 1.2. Today, 1.2 traffic is mostly legacy, and the arguments have shifted to certificate lifetimes and automation. Certificate authorities have been cutting maximum validity periods. A certificate that was valid for two years in 2020 may now be capped at a fraction of that, which turns manual renewal into an outage generator.

This is the quiet change of the last five years. The algorithm did not get harder to break. The operational schedule got tighter. Teams that automated renewal early barely noticed. Teams that did not have spent the intervening years firefighting.

Myth: Hashing is boring and settled#

Fact: Hashing is where the cost of migration shows up.

Password hashing moved from fast general purpose hashes to memory-hard functions, and the guidance has become firmer about parameters. Five years ago, many production systems still used a single round of SHA-256 with a salt. That pattern is now widely treated as inadequate, not because SHA-256 broke, but because the cost of guessing passwords on rented GPUs fell.

The harder problem is data at rest. Content addressed storage, code signing, and package registries depend on hash digests. Changing the digest function means revalidating every artifact that references it. A team that started in 2020 with a clean package registry can migrate in a quarter. A team with a decade of cached artifacts and pinned digests is looking at a multi year project. The algorithm choice is easy. The data archaeology is not.

Myth: Quantum computing is a distant problem#

Fact: The migration clock started years ago.

Harvest now, decrypt later is not a thought experiment. Anyone who captured encrypted traffic in 2020 and kept it is waiting for a future machine. The response from standards bodies has been to publish post-quantum algorithms and to recommend hybrid modes so that a break in either component does not sink the whole handshake.

Five years ago, this was a slide in a conference talk. Today it is a procurement question. Vendors get asked whether their hardware security modules support the new algorithms, and the honest answer is often not yet. The gap between the standard and the firmware update is where the next round of incidents will come from.

Myth: Longer keys mean stronger security#

Fact: Key management is the weak link, and it has not improved much.

Moving from RSA-2048 to RSA-4096 does little if the private key sits in a file with broad read permissions. The shift to elliptic curve keys and short lifetimes helps, but it also multiplies the number of keys in play. Five years ago, a mid sized company might have tracked a few dozen certificates. Today the same company may have thousands, issued by automation, with owners who left the company.

Key rotation is the recurring theme. The tools improved. Hardware security modules are more common. Cloud key services handle more of the plumbing. But the failure mode has not changed: a key that nobody knows about, in a system nobody owns, that expires on a Saturday.

What to check this quarter#

Start with the inventory. Count the certificates and keys in production, including the ones issued by automation. Compare that count to what the monitoring system knows about. The difference is your risk.

Next, look at the handshake. Confirm which key exchange groups your edge accepts and whether the load balancer in front of it can pass the larger post-quantum messages. Test with real clients, not just a browser.

Finally, check the hashes. List every place a digest is stored or pinned, from package manifests to signed documents. The migration will be slow, but the list itself is the first deliverable, and it is one that most teams could not have produced in 2020 either.

More articles