What’s the difference between supervised and adaptive spam filtering?

Supervised spam filtering uses labeled training data where humans mark examples as spam or legitimate, training models to recognize similar patterns. Adaptive filtering continuously learns from ongoing user feedback, adjusting to new patterns without explicit labeling of training data.

Supervised models establish baseline classification based on historical data. They excel at recognizing known spam patterns but may initially miss novel attacks. Retraining with new labeled data improves performance but requires human curation effort.

Adaptive systems learn from user actions in real time: spam button clicks, rescues from spam folders, and engagement patterns all inform the model. This enables faster response to new spam tactics but can also cause filtering changes based on individual user behavior that may not generalize well.