Algorithmic decision-making is the use of software rules, machine learning models, or scoring systems to produce or influence a choice. Instead of a person deciding directly, the system ranks options, flags risk, approves access, recommends an action, or sets a priority for review. In cyber security, this is common in fraud detection, spam filtering, identity verification, endpoint protection, and security triage.
It matters because automated decisions can be fast and consistent, but they can also be opaque and fragile. Attackers may exploit weak training data, poison inputs, or adversarially manipulate features to change a model’s output. Defenders need logging, explainability, access controls, human review for high-impact cases, and monitoring for drift or abuse. When a system decides who is trusted, blocked, or investigated, understanding how that decision is made becomes part of security itself.