Skip to content

3.6 Social Engineering & Process Attacks


Overview

These attacks exploit human factors and process weaknesses rather than technology. They are particularly dangerous because the person presenting to the camera IS a live human — so liveness detection alone cannot stop them.


Attack Types

Coercion & Duress

Attack Description Detection Mitigation
Physical coercion Victim forced to complete liveness under threat Stress detection (micro-expressions, pupil dilation) — unreliable Duress codes, behavioral analytics, post-verification checks
Social engineering of victim Victim tricked into completing liveness ("verify your account") N/A — victim cooperates willingly Customer education, transaction confirmation via separate channel
Insider collusion Bank employee manipulates verification or overrides results N/A at technology level Dual-approval workflows, audit trails, insider threat monitoring

Identity Exploitation

Attack Description Why Liveness Can't Help Mitigation
Identical twins Twin passes liveness and face matching for sibling Twin IS live and looks like the target Secondary verification (OTP to registered phone), behavioral biometrics, knowledge-based verification
Lookalike Person with similar appearance attempts verification Person IS live High-precision face matching (tighter thresholds), document verification
Willing account mule Legitimate person knowingly opens account for criminal use Person is genuinely who they claim to be Transaction monitoring, behavioral analytics, network analysis

Process Manipulation

Attack Description Mitigation
Fallback exploitation Deliberately fail digital liveness to trigger weaker manual process Ensure fallback processes are equally secure; don't make manual review "easier"
Session timing exploit Pass liveness, then manipulate data before face matching runs Atomic session processing; server-side orchestration; no client-controlled timing
Review queue manipulation Flood manual review queue with fraudulent applications to overwhelm reviewers Prioritized queuing, automated pre-screening, reviewer capacity planning

Key Insight

Technology Alone Is Insufficient

Social engineering attacks require process-level defenses: dual approvals, audit trails, behavioral monitoring, customer education, and insider threat programs. No liveness system, no matter how advanced, can detect a willing participant or an insider.


Next: Adversarial ML Attacks →