AI & Machine Learning
31 terms in this topic area.
Adversarial Machine Learning
Adversarial machine learning covers evasion, model extraction and membership inference. See how attackers fool AI models and how to defend them.
AI Backup and Model Recovery
AI backup and model recovery: what ML systems need backed up and how to check integrity. Plus clean rollback after ransomware or data poisoning.
AI Certification
AI certification covers ISO/IEC 42001, related standards and the BSI AIC4 catalogue. See how they relate to conformity assessment under the EU AI Act.
AI Data Protection Impact Assessment (AI DPIA)
An AI DPIA is a data protection impact assessment for AI systems. See when it is mandatory, which risks it covers and how it fits the AI Act.
AI Forensics
AI forensics uses AI to speed up digital investigations. It also examines incidents on AI systems: artefacts, methods and limits.
AI Governance for Managing Directors
AI governance for managing directors: duty of care and liability, AI literacy under Art. 4 AI Act and solid decision papers for AI projects.
AI Governance Framework
An AI governance framework sets roles, policies, an AI inventory and a risk process. Based on ISO/IEC 42001 and the NIST AI RMF, for mid-sized firms.
AI Incident Response Playbook
An AI incident response playbook is a practical emergency plan. It covers AI incidents, first steps, containment and reporting under GDPR, NIS2 and AI Act.
AI Monitoring and Anomaly Detection
AI monitoring and anomaly detection use machine learning to spot unusual behaviour in networks, endpoints and accounts. Limits and first steps.
AI Penetration Testing
AI penetration testing checks LLM applications and ML models for prompt injection, data leakage and manipulation. Methodology, scope and tools.
AI Phishing Detection
AI phishing detection uses machine learning to assess emails, links and sender behaviour. See its limits and what companies should consider.
AI Poisoning
AI poisoning uses manipulated training data, backdoors or poisoned RAG sources to compromise AI models. Here is which controls actually help.
AI Red Teaming
AI red teaming uses offensive tests to find prompt injection, data leakage and unsafe behaviour in AI applications. Process, methods and tools.
AI Risk Management
AI risk management for SMEs: list your AI use and assess the risks. Use the AI Act, ISO/IEC 23894 and NIST AI RMF, then add them to your ISMS.
AI Security in Critical Infrastructure
AI security in critical infrastructure: operator duties under the BSI Act and NIS2. Plus high-risk AI under the AI Act and practical controls.
AI Supply Chain Attack
An AI supply chain attack uses tampered models, datasets or Python packages. See how they reach your company and how to secure their origin and integrity.
AI-Based Biometric Authentication
AI-based biometric authentication verifies faces, voices or fingerprints. See key attacks such as deepfakes and what GDPR and the AI Act require.
Algorithmic Accountability
Algorithmic accountability asks who answers for automated decisions. See what the GDPR and EU AI Act require and how companies prove responsibility.
Algorithmic Bias
Algorithmic bias leads to unfair AI decisions. See causes, fairness metrics, countermeasures and duties under German equal treatment law, GDPR and AI Act.
Chatbot Security
Chatbot security protects LLM-based AI assistants in your business. See how to detect and contain prompt injection, data leakage and excessive permissions.
Edge AI
Edge AI runs models on IoT devices. See risks such as model extraction, physical tampering and insecure updates, and how hardening and secure boot help.
EU AI Act Compliance
EU AI Act compliance for companies in Germany: risk classes, provider and deployer duties. Plus deadlines from 2025, fines and first steps.
Explainable AI (XAI)
Explainable AI (XAI) makes AI decisions traceable. See methods such as SHAP, LIME and counterfactuals and their role under the EU AI Act and GDPR.
Federated Learning
Federated learning trains AI without pooling raw data. See which attacks it faces, from poisoning to gradient inversion, and how to secure it.
GDPR-Compliant AI Systems
GDPR-compliant AI systems need a legal basis and transparency. See Article 22, impact assessments and data subject rights explained for businesses.
Homomorphic Encryption
Homomorphic encryption lets AI compute on encrypted data. See PHE, SHE and FHE, practical use cases, limits and how to get started.
MLOps Security
MLOps security protects data, training environments, model registries and the deployment of ML models. Concrete measures for each pipeline stage.
NIS2 Compliance for AI Systems
NIS2 compliance for AI systems under Germany's revised BSIG. See risk management, supply chain, incident reporting and management duties.
Privacy-Preserving Machine Learning
Privacy-preserving machine learning (PPML) uses differential privacy, federated learning, homomorphic encryption and confidential computing.
Quantum-Safe AI
Quantum-safe AI uses post-quantum cryptography (ML-KEM, ML-DSA). It protects training data, model weights and AI interfaces against harvest now, decrypt later.
Synthetic Data
Synthetic data replaces real records in testing and AI training. See when it is truly anonymous, which risks remain and how to evaluate it.