A Logistics Firm's Crypto Choice: Keep RSA or Add Post-Quantum Keys

Archi4 min read

cryptographytlssecuritymigration

Contents
  1. The Lock Decision
  2. Three Jobs, One Handshake
  3. The Three Options on the Table
  4. What Broke in the Test Lab
  5. The Cost of a Downgrade
  6. What Other Teams Should Copy

The Lock Decision#

A mid-sized freight software company had a simple question with an awkward answer: what should it do about encryption that a future quantum computer could break? The company moves customs documents and invoices between carriers, brokers, and banks. It keeps records for seven years. That retention rule turned a distant research problem into a current business decision.

The team used TLS 1.3 for every API call. Certificates were RSA. Session keys came from elliptic curve Diffie-Hellman, usually X25519. Both are standard. Both depend on math that a large quantum computer could solve. Nobody has built that machine. The risk is different: an attacker can record encrypted traffic today and decrypt it later. Security engineers call this harvest now, decrypt later. For a company with seven years of retained data, the window matters.

Three Jobs, One Handshake#

Public key cryptography does three jobs in a TLS connection. A signature proves the server is who it claims. A key exchange lets two sides agree on a shared secret. Symmetric encryption, usually AES, protects the actual data. Quantum computers threaten the first two jobs, not AES in the same way. A 256-bit AES key still has a large margin.

RSA and elliptic curve cryptography are efficient, but they rely on factoring and discrete logarithms. A quantum machine running Shor's algorithm could break both. Grover's algorithm speeds up brute force against symmetric keys but does not break AES-256. That is why the practical response centers on signatures and key exchange, not on replacing AES.

The Three Options on the Table#

The first option was to wait. Keep RSA certificates and X25519 key exchange. The cost was zero this quarter. The risk was a longer exposure window if traffic is recorded now.

The second option was to swap in post-quantum signatures immediately. NIST has standardized algorithms for this, including ML-DSA. The company's certificate authority could issue new certificates. But every client, load balancer, and partner integration would need to trust them. Certificate chains would grow. Some old Java and mobile clients would fail. The operations team estimated months of partner coordination.

The third option was a hybrid key exchange. Keep existing RSA certificates for signatures. Add a post-quantum key encapsulation method alongside X25519 during the TLS handshake. The client and server mix two secrets. An attacker must break both to recover the session key. The company chose this route. It did not fix signatures, but it addressed the recorded-traffic risk for new sessions.

What Broke in the Test Lab#

The first test passed. The second test failed on an older Android SDK. The handshake message grew by several hundred bytes. A middlebox in one partner's network rejected the larger ClientHello. A load balancer had a fixed buffer that the vendor had never documented. None of these failures were cryptographic. They were plumbing.

The team split the rollout. Internal service-to-service traffic moved first. Admin consoles came next. The public API got a feature flag and a dashboard that counted hybrid handshakes by client version. Failed handshakes fell back only where the business allowed it. The security lead wanted a hard cutover. The API lead wanted six months. They settled on a deadline with a monthly review.

The Cost of a Downgrade#

A hybrid handshake is negotiated. If a client does not offer the new group, a server may choose the old one. That is compatibility. It is also a downgrade path. For passive recording, a forced downgrade requires an active attacker. The team treated that as a separate risk. It made hybrid mandatory for internal traffic and for partners that had completed testing. For public clients, it watched the fallback rate and set a date to require hybrid. The date was a business decision, not a cryptographic one.

The finance team asked a fair question: what happens if quantum computers arrive later than expected? The answer was that hybrid key exchange costs CPU and bytes, not a rewrite. The company could turn it off. Certificates are the harder migration. The team started an inventory of every RSA and ECDSA certificate, including test systems and vendor appliances.

What Other Teams Should Copy#

Separate the two problems. Key exchange protects each session. Signatures protect identity and software updates. A company can fix one before the other.

Put the handshake on a scale. Test with the oldest phone, the strangest proxy, and the busiest load balancer. Measure handshake size, CPU, and failure modes.

Make fallback visible. A counter is not a security control, but it tells you whether clients are ready. A silent fallback is a hidden deadline.

Keep the inventory boring. A spreadsheet with owner, system, algorithm, and expiration date beats a slide deck. Cryptography decisions age badly when nobody knows where the keys live.

The freight company did not solve quantum risk. It bought time with a hybrid handshake and a dated plan for signatures. That is the shape of most cryptography decisions: a narrow choice, made with incomplete information, and a list of what still needs to change.

More articles