IS audit planning · Lesson 3

Control design and audit criteria

4 min read

Lesson objectives

  • Translate a defined risk into a control objective and testable design and operation
Reading tools

When someone says “We have a control,” first ask which failure it prevents or detects. The objective connects risk to action. In a payments system, one person might change a beneficiary bank account without review. A control could prevent an unauthorized change from reaching production or detect it quickly enough for correction. Starting with a generic checklist of technical settings can prove that a feature exists while missing the risk altogether.

Controls play different roles. A preventive control blocks an event, such as separating creation from approval rights. A detective control signals that an event occurred, such as a daily reconciliation of beneficiary changes against approvals. A corrective control limits harm and restores an approved state, such as freezing an improper payment and recovering the correct record. Calling a control “preventive” does not establish that it works. Consider who can bypass it, which channels it covers, when it operates, and whether its input data are reliable. Strong detection may partly compensate for weak prevention in a particular setting, but a warning after funds have left cannot erase that exposure.

Build a four-part test chain: risk, criterion, evidence, and permissible conclusion. The hypothetical organization's criterion here is that an independent person reviews every beneficiary-account change before the first payment to the new account. Design testing needs an entitlement model showing that a creator cannot approve their own change, a description of the workflow, and a trustworthy timestamp source. Operating testing needs the full population of changes over a stated period and a link from each change to its reviewer and first payment time. A manager interview can explain the process but does not establish that it operated in every instance. One screenshot shows an interface at one moment; it does not prove the back-end separation rule or its behavior throughout a quarter.

Try to disprove your conclusion. Does the population cover every change channel? Can a privileged operator alter a beneficiary outside the log? Are both timestamps in the same time zone? These are consequential questions: a bad assumption could reverse the finding. NIST SP 800-53A offers adaptable assessment procedures, but it does not impose this case's exact policy on every organization. Record the organization's criterion and its source, then limit the conclusion to the month, control, and channels tested.

Lesson summary

Try to disprove your conclusion. Does the population cover every change channel? Can a privileged operator alter a beneficiary outside the log? Are both timestamps in the same time zone? These are consequential questions: a bad assumption could reverse the finding. NIST SP 800-53A offers adaptable a