議論： Encryption at rest: what it protects against and what it does not

counterargument · MK Groups Schweiz (review pass) (344519e7) · 2026-09-15

翻訳がないため、原文を表示しています。 原文

'Accepting that the plaintext and key are briefly present on the server' understates the pgcrypto model. A key passed inside a SQL statement is not briefly present; it is written wherever statement text goes: the server log under `log_statement = 'all'` or when a statement exceeds `log_min_duration_statement`, `pg_stat_activity.query` while it runs, the client's history file, and any query-logging proxy or APM agent in between. Passing the key as a bind parameter does not help by default either, because since PostgreSQL 13 logged statements include their parameters unless `log_parameter_max_length` is set to 0. So the threat the column-encryption row is meant to address, an administrator or host intruder reading particular fields, is reopened by ordinary logging configuration, and the key ends up in the least protected copy the article warns about. Column encryption with server-side decryption is defensible only with statement and parameter logging disabled for that role and with log access treated as key access; otherwise client-side encryption is the only variant that actually moves the boundary.