Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Deterministic encryption produces the same ciphertext for the same plaintext every time, which makes equality search possible and leaks far more than most teams expect.

What Are the Risks of Deterministic Encryption?
Encrypt the surname Smith twice with a randomised scheme and you get two unrelated ciphertexts. Encrypt it twice deterministically and you get the same value both times. That single property is what makes encrypted database columns searchable, joinable, and indexable — and it is also what gives an attacker a map of your data without ever recovering a key.
Teams usually arrive at deterministic encryption for a practical reason: the application needs WHERE email = ? to work, or two encrypted tables need to be joined, or a unique constraint has to hold. The trade-off is real, but it is often accepted without anyone writing down what it costs.
Deterministic ciphertext preserves the equality relation of the plaintext, and equality is enough to reconstruct a great deal.
The academic work is blunt about this. Naveed, Kamara, and Wright demonstrated in 2015 that inference attacks against property-preserving encrypted databases recovered a large proportion of attributes from real electronic health records, including mortality and sex, using nothing but public statistics and the ciphertext. Order-preserving encryption, which also leaks sort order, fared even worse.
| Setting | Why deterministic | Exposure |
|---|---|---|
| SQL Server Always Encrypted, deterministic mode | Supports equality search, joins, grouping | Documented by Microsoft as leaking patterns in low-entropy columns |
| Application-side column encryption | Needed for lookups and unique keys | Usually the same key across all rows and environments |
| Tokenisation with a fixed mapping | Preserves referential integrity | Equivalent to deterministic encryption unless the vault randomises |
| Key wrapping and key derivation | Determinism is deliberate and safe | Inputs are high-entropy keys, not guessable values |
That last row matters. Determinism is not wrong in itself; it is wrong when applied to low-entropy, guessable, or repeated plaintext. AES-SIV, specified in RFC 5297, is a deterministic authenticated mode designed precisely for wrapping keys, and it is the right tool there.
Deterministically encrypted personal data is still personal data. The ICO treats pseudonymisation as a security measure, not as anonymisation, and data that can be re-linked to an individual — which deterministic ciphertext plainly can, by anyone holding a matching record — remains in scope for UK GDPR. Describing such a column as anonymised in a privacy notice or a data protection impact assessment is a misstatement worth catching before a regulator does.
Encryption may still reduce the risk rating of a breach under Article 34, but only if the exposed data could not reasonably be linked to individuals. A frequency-analysable column of diagnoses does not clear that bar.
Our infrastructure and architecture review services examine exactly these design decisions, because encrypted-at-rest is frequently recorded as a completed control when what was actually deployed leaks most of the information it was meant to hide.
No. It is the correct choice for key wrapping, where the plaintext is a random key with no guessable structure, and AES-SIV exists for that purpose. The problem arises when it is applied to low-entropy business data such as names, postcodes, statuses, or dates, where repetition and frequency carry most of the meaning.
You normalise the plaintext, compute an HMAC with a key reserved for indexing, truncate the result to a few bytes, and store that alongside randomised ciphertext. Queries hash the search term the same way and filter on the index, then decrypt candidates and compare exactly. Truncation causes deliberate collisions, which blurs frequency while keeping lookups fast.
It breaks equality, joins, sorting, and unique constraints on the encrypted column, which is why teams reach for determinism in the first place. The usual answer is to keep the encryption randomised and add a separate blind index or a search service for the specific queries you need, rather than weakening the encryption for every row.
Yes, where you or anyone else can reverse the link to an individual. The ICO treats encryption and pseudonymisation as security measures rather than anonymisation. Deterministic ciphertext is especially clearly in scope, because matching a known value against the stored ciphertext re-identifies records without any key at all.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.