Please enable JavaScript.
Coggle requires JavaScript to display documents.
AI & ML - Coggle Diagram
AI & ML
Privacy and Security Challenges
Data Leakage Risks
Scenario 1: Medical AI model inadvertently reveals patient data patterns
Option 1: Irreversible anonymization (recommended)
Option 2: Tokenization
Scenario 2: Financial prediction model exposes account holder behavior
Option 1: Federated learning (recommended)
Option 2: Data silos
Theory: Rights Based Ethics (individual privacy as a right)
Algorithmic Bias
Scenario 1: Facial Recognition fails for dark-skinned users
Option 1: Diverse training data (recommended)
Option 2: Algorithmic fairness audits
Theory: Rawlsian Justice (veil of ignorance)
Scenario 2: Loan AI disadvantages marginalized groups
Option 1: Explainable AI (recommended)
Option 2: Unmodified deployment
Theory: Critical Race Theory (systemic bias awareness)
Cybersecurity Advantages
Threat Detection
Scenario 1: Zero-day exploit identification
Option 1: Full automation (recommended for speed)
Option 2: Human-AI hybrid
Theory: Consequentialism (best outcomes matter most)
Scenario 2: Phishing pattern prediction
Option 1: Real-time alerts (recommended)
Option 2: Delayed human review
Theory: Pragmatism
Data Protection
Scenario 1: Hospital network anomaly detection
Option 1: AI-managed encryption (recommended)
Option 2: Manual key rotation
Theory: Virtue Ethics (reliability as a virtue)
Scenario 2: Automated regulatory compliance
Option 1: Continuous monitoring (recommended)
Option 2: Quarterly audits
Theory: Rule Utilitarianism (systemic rule-following)
Cybersecurity Risks
Adversarial Attacks
Scenario 1: Malicious noise fools autonomous vehicles
Option 1: Robust adversarial training (recommended)
Option 2: Basic hardening
Theory: Deontology (duty to protect users)
Scenario 2: Poisoned Training data manipulates fraud detection
Option 1: Data provenance tracking (recommended)
Option 2: trusted third-party datasets
Theory: Social Contract Theory (mutual system of trust)
Model Inversion
Scenario 1: Attackers reconstruct medical training data
Option 1: Differential privacy (recommended)
Option 2: output filtering
Theory: Libertarianism (self-ownership of data)
Scenario 2: Membership inference exposes participants
Option 1: k-anonymity protection (recommended)
Option 2: no modification
Theory: Feminist ethics (power imbalance awareness)
Ethical Challenges
Surveillance Ethics
Scenario 1: Workplace productivity monitoring
Option 1: Strict usage policies (recommended)
Option 2: unrestricted monitoring
Theory: Kantian Ethics (respect for human dignity)
Scenario 2: Public facial recognition for security
Option 1: Opt-in consent (recommended)
Option 2: mandatory scanning
Theory: Communitarianism (community vs individual rights
Responsibility Gaps
Scenario 1: Autonomous pentesting causes system crash
Option 1: Human-in-the-loop (recommended)
Option 2: Full AI autonomy
Theory: Virtue Ethics (accountability as a virtue)
Scenario 2: AI-generated security report errors
Option 1: Mandatory human review (recommended)
Option 2: unverified AI reports
Theory: Care Ethics (emphasis on human oversight)