Optical-layer encryption has become a serious option. Modern coherent transponders can encrypt an entire wavelength with AES-256 at line rate.
If the optical layer is already quantum-safe, why would you still need Layer 2 or Layer 3 encryption on top of it?
The short answer is that Layer 1 encryption answers a narrower question than most organizations think they’ve answered. It protects a wavelength between two optical endpoints. It doesn’t protect your data everywhere that data travels, it doesn’t separate one kind of traffic from another, and it may not be under your control at all, depending on how the service is delivered.
Whether that gap matters depends entirely on the link. This page covers where Layer 1 alone is enough, where it isn’t, and how to tell the difference.
Layer 1 (optical or OTN) encryption works on the transport signal itself. The transponder at one end encrypts everything carried on the wavelength or OTN circuit; the transponder at the other end decrypts it. Between those two points, anyone tapping the fiber sees ciphertext. Outside those two points, the traffic is in the clear.
That boundary is the key. Three properties follow from it:
None of these are flaws. They’re what Layer 1 is. But each one defines a situation where Layer 2 or Layer 3 encryption is still doing work that Layer 1 cannot.
1. The path ends where the fiber ends. Very few data flows stay on a single optical segment from source to destination. A replication stream might cross encrypted dark fiber between two data centers and then continue over a routed WAN to a third site, into a cloud region, or across a partner connection. Every routed, MPLS, internet, 4G/5G or satellite segment beyond the optical domain is unprotected by Layer 1. Layer 3 encryption (IPsec) is what follows the data through routed infrastructure. Layer 2 encryption (MACsec) is what follows data across switched, carrier-Ethernet and metro segments that aren’t a single wavelength. Harvest-now-decrypt-later adversaries don’t need to tap the best-protected segment but only the weakest.
2. The last metres are in plaintext. Even on a fully encrypted wavelength, traffic is in the clear between the transponder and the switch or router it connects to, including patch panels, meet-me rooms, colocation cross-connects, carrier demarcation points, shared cages. These are exactly the places where physical access is hardest to guarantee and easiest to overlook. Encrypting at Layer 2 or Layer 3 at the network edge, inside your own trusted boundary, means the traffic is already ciphertext before it ever reaches the optical gear, so a cross-connect in someone else’s building carries nothing readable.
3. Who holds the keys? With owned transponders, you control the keys. With a managed or encrypted-wavelength service, the encryption may be configured, operated and monitored by the provider, and key generation or escrow may sit partly or wholly on their side. For regulated sectors that’s even more a governance question than a technical one: DORA’s third-party ICT risk requirements, NIS2 supply-chain obligations, and sovereignty and classified-data rules all ask who can access the keys protecting your data. A customer-operated Layer 2 or Layer 3 encryption layer keeps keys, policy and audit trail under your own control, regardless of who runs the transport underneath.
4. One wavelength, one policy. Layer 1 can’t tell one application, tenant, VLAN, or classification level from another. Modern security architectures increasingly require exactly that separation: production apart from management, OT apart from IT, one business unit or customer apart from another, restricted data apart from everything else. Layer 2 and Layer 3 encryption can apply distinct keys and tunnels per segment, per VLAN, per site pair or per traffic class. This ensures a compromise or misconfiguration in one segment doesn’t expose all of them, and so segmentation policy survives the trip across a shared transport.
5. A single layer is a single point of cryptographic failure. Every cryptographic implementation can fail, for example, due to an algorithm weakness, an implementation bug, a key-management error, or a misconfigured device. Layered encryption from independent implementations is an established principle for high-assurance data precisely because of this.
National-security frameworks such as the NSA’s Commercial Solutions for Classified (CSfC) program require two independent encryption layers for classified data in transit. This matters doubly where the optical layer’s quantum-safety relies on QKD: European security agencies (ANSSI, BSI, the Netherlands’ NLNCSA and Sweden’s national communications security authority, in a joint January 2024 position paper) concluded that QKD’s security maturity is not yet sufficient and recommended prioritizing post-quantum cryptography. Moreover, the NSA reaffirmed in October 2026 that it does not recommend QKD for national security systems. A PQC-based Layer 2 or Layer 3 layer gives you an independently quantum-safe second line, built on a different mechanism.
There’s a sixth, softer gap worth naming: lifecycle. Optical platforms are refreshed on transport timescales — often a decade or more — while post-quantum standards concerning new algorithms, revised parameters, and finalization of FN-DSA are still moving or being finalized. A crypto-agile encryption layer that updates in software keeps your quantum-safe posture current even when the optical estate isn’t due for renewal.
Layer 1 on its own can be a defensible choice when all of the following are true:
If any one of these is false, Layer 1 is still worth keeping — but it isn’t the whole answer.
| Situation | Why Layer 1 alone falls short | What to add |
| Data continues beyond the optical segment (WAN, MPLS, cloud, internet, 5G, satellite) | Protection ends at the transponder | Layer 3 encryption end to end |
| Metro or carrier-Ethernet segments that aren’t a single wavelength | No optical encryption on those hops | Layer 2 encryption across the switched path |
| Colocation, meet-me rooms, third-party cross-connects | Plaintext between transponder and your edge | Layer 2 or 3 at your own network edge |
| Multiple tenants, business units, VLANs or classification levels on one wavelength | One key and one policy for everything | Layer 2 per-VLAN/segment or Layer 3 per-tunnel separation |
| Classified, defense, or high-assurance data | Single-layer encryption, often QKD-dependent | Independent PQC-based Layer 2 or 3 as a second layer |
| Long-lived optical estate, fast-moving PQC standards | Crypto updates tied to transport refresh cycles | Crypto-agile Layer 2/3 that updates in software |
SSH’s NQX is designed to work alongside optical encryption. NQX encrypts at Layer 2 and Layer 3 simultaneously and from the same appliance. It uses hybrid post-quantum key agreement (FrodoKEM and Kyber/ML-KEM) alongside classical algorithms, at wire speed up to 100 Gbps.
NQX runs over any transport, encrypted wavelengths included, so it adds an independent, customer-controlled quantum-safe layer that follows your data beyond the fiber, keeps segments cryptographically separate, and keeps keys and audit evidence in your hands. Its software-defined crypto engine takes new algorithms as updates, not hardware refreshes.
→ Explore NQX quantum-safe network encryption
→ Choosing between layers: Layer 2 vs. Layer 3 quantum-safe encryption
→ Compare approaches: PQC vs. QKD vs. pre-shared keys
Is optical (Layer 1) encryption quantum-safe? It can be. Wavelength encryption uses AES-256, which holds up against quantum attacks. What decides it is how the keys are agreed. Look for post-quantum or hybrid key exchange, or QKD backed by post-quantum authentication. If keys are still exchanged with classical RSA or elliptic-curve algorithms, the link is exposed to harvest-now-decrypt-later attacks.
If our wavelengths are encrypted, do we still need Layer 2 or Layer 3 encryption? You do if your data leaves the optical segment, if the carrier holds the keys, if different traffic classes share one wavelength, or if your rules require two independent layers. If none of these apply, Layer 1 alone can be enough.
Can Layer 2 or Layer 3 encryption run over an encrypted wavelength? Yes. Network encryptors work independently of the transport underneath, so they run over encrypted or unencrypted wavelengths, dark fiber, carrier Ethernet or IP networks. The two layers use independent keys and implementations. That is exactly what defense-in-depth frameworks ask for.
Isn't encrypting twice wasteful? Not for high-value data. Two independent layers remove the single point of cryptographic failure. The NSA's Commercial Solutions for Classified program requires two layers for classified data in transit. For low-risk links that stay on one owned optical segment, one layer may be enough. Decide link by link.
Who holds the keys on a managed encrypted wavelength service? Often the carrier, at least in part. If the provider configures and operates the encryption, it may also generate or escrow keys. DORA third-party rules, NIS2 supply-chain duties and classified-data regulations all ask who can access your keys. A customer-operated Layer 2 or 3 layer keeps the keys in your hands.
Does QKD on the optical layer make a link quantum-safe? Not on its own, according to most national security agencies. They recommend prioritizing post-quantum cryptography. A PQC-based Layer 2 or Layer 3 layer adds an independent quantum-safe mechanism on top.