Affect recognition is the automated inference of emotional state from signals such as facial expression, tone of voice, speech patterns, posture, or other behavior. It does not measure feelings directly; it classifies observable cues and produces a best-effort guess, which can be wrong or overconfident. In cyber security, that matters because these systems process sensitive biometric and behavioral data, often inside products that also handle identity checks, recordings, and user profiling.
Attackers can abuse affect recognition to make social engineering more persuasive, adapting scam scripts to a victim’s apparent stress, hesitation, or trust. Defenders may use it for safety moderation, customer support triage, or fraud detection, but the output should be treated as a probabilistic signal, not fact. Good security practice requires clear consent, limited retention, strict vendor controls, and caution when emotion inference is combined with face or voice analysis.