A complete walkthrough of the TryHackMe room AI System Reconnaissance (Module 2), scanning ML services, fingerprinting Triton, MLflow, and Qdrant, extracting model metadata, and mapping the Cyphira AI attack surface.
A complete walkthrough of the TryHackMe room AI Threat Modelling (Module 2), adapting STRIDE for AI systems, mapping MITRE ATLAS techniques, applying the OWASP LLM Top 10 to enterprise architecture, and completing MegaCorp's threat model assessment.
A complete walkthrough of the TryHackMe room LLM Security (Module 2), analyzing data-based, model-based, system-based, and user-based attack surfaces against LLMs.
A complete walkthrough of the TryHackMe room Securing AI Systems (Module 2), mapping AI architectures, trust boundaries, OWASP LLM Top 10 vulnerabilities, and conducting a pre-deployment audit of TryAssist.
A complete walkthrough of the TryHackMe room AI Forensics, exploring how machine learning powers digital forensics, legal and ethical implications, and an interactive breach investigation using scikit-learn.
A complete walkthrough of the TryHackMe room Prompt Engineering, covering LLM tokenisation, sampling parameters, the four pillars of prompt design, instruction hierarchies, and the PromptSec challenge.
A complete walkthrough of the TryHackMe room AI Models & Data, exploring training data provenance, model building risks, the fine-tuning inheritance problem, model cards, and HuggingFace supply chain audits.
A complete walkthrough of the TryHackMe room AI Security Threats, covering model vulnerabilities, AI-enhanced attacks, defensive AI capabilities, and secure adoption standards.
A complete walkthrough of the TryHackMe room The Building Blocks of AI, covering the ML lifecycle, learning paradigms, neural networks, and transformer-based LLMs.
A hands-on lab demonstrating how an over-broad kms:Decrypt and secretsmanager:GetSecretValue grant on a single low-privilege role exposes secrets and ciphertext belonging to completely unrelated applications and how a KMS grant becomes a persistence mechanism that survives IAM credential rotation.