Course curriculum

  • 1

    Module 1: Foundations of AI and AI Security

    • Evolution of AI and Security Challenges

    • Why AI Systems Need Dedicated Security Models

    • Differences Between Traditional AppSec and AI Security

    • AI Security Terminology and Core Concepts

    • Machine Learning, Deep Learning, and Generative AI

    • Supervised, Unsupervised, and Reinforcement Learning

    • Model Training, Evaluation, and Deployment Basics

    • AI Use Cases in Enterprises and AI Failure Modes

    • CIA Triad and Core Security Principles

    • Threat Modeling Basics

    • Secure Software Development Lifecycle (SSDLC)

    • Identity, Access Management, Cryptography Basics

    • Network and Cloud Security Fundamentals
  • 2

    Module 2: AI Architecture, MLOps, and Data Security

    • End-to-End AI System Architecture

    • Data Pipelines, Feature Stores, and Data Flows

    • Model Lifecycle Management

    • CI/CD for ML (MLOps)

    • Risks in AI Deployment Pipelines

    • Data Classification and Sensitivity

    • Data Poisoning and Data Tampering Attacks

    • Secure Data Storage and Access Control

    • Data Lineage and Provenance

    • Data Retention and Secure Deletion
  • 3

    Module 3: Privacy, Governance, and Responsible AI

    • Privacy Risks in AI Systems

    • Anonymization and Pseudonymization Techniques

    • Consent, Purpose Limitation, and Lawful Processing

    • Privacy by Design in AI Systems

    • AI Risk Management Frameworks

    • Policies, Standards, and Controls for AI

    • Regulatory Landscape

    • Audit, Evidence, and Compliance Reporting

    • Ethical and Responsible AI Governance

    • Relationship Between Security, Safety, and Ethics

    • Building Trustworthy AI Systems
  • 4

    Module 4: Threat Modeling and Adversarial ML

    • AI-Specific Threat Modeling Approaches

    • Attack Surfaces in ML Pipelines

    • STRIDE and MITRE ATLAS for AI

    • Abuse Case and Misuse Case Design

    • Risk Scoring and Prioritization

    • Adversarial Examples and Evasion Attacks

    • Poisoning and Backdoor Attacks

    • Model Inversion and Membership Inference

    • Model Extraction and Theft

    • Defense Strategies and Their Limitations
  • 5

    Module 5: Secure AI Development and Operations

    • Security Requirements for AI Projects

    • Secure Design and Architecture Reviews

    • Trusted Datasets and Data Validation

    • Secure Training Environments

    • Reproducibility and Integrity Checks

    • Preventing Backdoors and Trojans

    • Secure Evaluation and Benchmarking

    • Secure Model Serving Architectures

    • API Security for AI Services

    • Access Control and Rate Limiting

    • Model Integrity Verification

    • Runtime Monitoring and Drift Detection

    • Secure Deployment and Operations

    • Continuous Improvement and Maturity Models
  • 6

    Module 6: Generative AI, LLM, and Prompt Security

    • LLM Architectures and Risk Profile

    • Prompt Injection and Jailbreak Attacks

    • Data Leakage and Training Data Exposure

    • Hallucinations, Abuse, and Safety Risks

    • Guardrails, Filters, and Policy Enforcement

    • Prompt Design from a Security Perspective

    • Input Validation and Sanitization

    • Context Isolation and Tool Security

    • Reducing Prompt-Based Attack Surface
  • 7

    Module 7: AI Infrastructure and Cloud Security

    • Securing GPUs, TPUs, and AI Accelerators

    • Secrets, Keys, and Credential Management

    • Network Segmentation for AI Workloads

    • Supply Chain Security for AI Platforms

    • Shared Responsibility Model for AI

    • Securing Managed AI Services

    • Storage, Compute, and IAM Hardening

    • Multi-Tenant Risks and Isolation
  • 8

    Module 8: AI Security Operations, Detection, and Testing

    • What to Log in AI Systems?

    • Security Telemetry and SIEM Integration

    • Detecting Model Abuse and Anomalies

    • Drift, Degradation, and Data Shifts

    • Alerting and Automated Response

    • AI-Specific Incident Types

    • Containment of Model and Data Breaches

    • Forensic Analysis of Data Pipelines and Models

    • Recovery, Re-Training, and Re-Deployment

    • Post-Incident Reviews and Hardening

    • Adversarial Testing of Models and Prompts

    • Penetration Testing AI APIs and Pipelines

    • Reporting, Metrics, and Risk Communication
  • 9

    Study Material

    • Study Material