Cybersecurity & Post-Quantum Security Engineer

An engineer is asked why the company cannot simply lengthen its RSA keys to stay ahead of future attacks. It is a reasonable question with an unreasonable answer: the security of RSA does not degrade gradually with key length against the attack that matters most. Understanding why requires knowing what a cryptographic system actually rests on, and that foundation is a specific mathematical assumption rather than a general notion of difficulty.

Two families, two jobs

Symmetric cryptography uses one shared key for both encryption and decryption. It is fast, well understood, and used for essentially all bulk data — AES is the standard example. Its structural problem is key distribution: two parties who have never met cannot agree a shared secret over a channel someone is listening to.

Asymmetric cryptography solves exactly that. Each party holds a mathematically linked key pair, publishes one half, and keeps the other. Anyone can encrypt to the public key; only the private key decrypts. The same structure supports digital signatures, where the private key produces a signature anyone can verify against the public key.

In practice the internet uses both together: asymmetric cryptography establishes a shared secret, and symmetric cryptography protects the data. TLS works this way, which is why an attack on the asymmetric layer compromises sessions whose bulk encryption is untouched.

What a hard problem actually means

RSA rests on the difficulty of factoring a large number into its prime components. Elliptic curve cryptography rests on the difficulty of the discrete logarithm problem over a curve. These problems share a critical property: they are believed hard for classical computers, based on decades of failed attempts to find efficient algorithms.

That phrasing carries the whole risk. Security here is not a proof, it is an absence of a known efficient attack. The assumption holds until someone finds one — and a quantum algorithm targeting exactly these problems is what changes the picture. This is why lengthening keys does not help against the relevant threat: it raises the classical cost while leaving the structural assumption intact.

Failure modes

Custom cryptography — implementing a scheme rather than using a reviewed library. Signal: cryptographic primitives written in-house with no external audit.

Algorithm agility absent — the choice hardcoded across a codebase. Signal: a migration estimate measured in quarters because the algorithm appears in dozens of places.

Key management neglected — strong algorithms with keys stored beside the data. Signal: keys in configuration files or repositories.

Deprecated primitives in place — algorithms broken years ago still in production paths. Signal: an inventory nobody has produced, so nobody knows.

Explaining this to stakeholders

On the two families: symmetric encryption is a shared padlock where both people already have the key. Asymmetric is a letterbox — anyone can drop something in, only the holder of the key can open it. The internet uses the letterbox to hand over the padlock key, then uses the padlock.

On hard problems: these systems are secure because nobody has found a fast way to break the underlying maths, not because it is proven impossible. That distinction is the entire risk.

On longer keys: a bigger key makes the existing attack slower. It does not help if someone finds a fundamentally faster method, which is the situation we are preparing for.