An objective function is the mathematical target a machine learning system tries to minimize or maximize during training. It can represent error rate, cost, latency, risk, or a weighted mix of several goals. The model learns by adjusting its parameters to improve that score, so the objective function effectively defines what “success” means.
In cyber security, objective functions matter because they shape both defenses and attacks. A defender may tune a detector to reduce false negatives, false positives, or overall risk, and a bad objective can make a system look accurate while missing dangerous behavior. Attackers also exploit objectives: malware may try to evade a classifier, phishing campaigns may optimize for click rates, and adversarial machine learning may search for inputs that force a model toward the wrong output. When the objective is narrow or mis-specified, the system can optimize the metric while failing the security goal. That is why security teams review training targets, thresholds, and post-processing rules as carefully as the model itself.