EXIN AI Security Professional study plan

Map the AI system attack surface by listing at least 10 distinct threat categories (data poisoning, adversarial input, model extraction, prompt injection, supply-chain compromise, insecure output handling, training-data leakage, model inversion, denial-of-service, agent/tool abuse) and writing a one-sentence definition of each in your own words

Explain the CIA triad as applied to ML pipelines, distinguishing confidentiality risks (training-data and model-weight exposure) from integrity risks (tampering, poisoning) and availability risks (resource exhaustion)

Diagram a reference ML lifecycle (data collection, preprocessing, training, evaluation, deployment, inference, monitoring, retirement) and annotate at least two security controls at every stage

Define the difference between traditional software vulnerabilities and ML-specific vulnerabilities, citing at least three concrete examples of each

Analyse prompt injection and jailbreak mechanics by categorising at least five attack patterns (direct instruction override, indirect injection via retrieved content, role-play escalation, encoding obfuscation, multi-turn drift) and note one mitigation per pattern

Evaluate defences for LLM applications: input/output filtering, privilege separation, least-privilege tool access, human-in-the-loop gating, and system-prompt hardening, stating the residual risk each leaves

Compare adversarial-example defences (adversarial training, input sanitisation, randomised smoothing, detection classifiers) on cost, robustness gain, and known limitations

Describe model extraction and model inversion attacks and explain which data-protection obligations they trigger

Explain the shared-responsibility split for AI across cloud provider, model provider, and deploying organisation, and identify which controls sit with whom

Outline data security controls for training and inference data: classification, minimisation, anonymisation and pseudonymisation limits, encryption at rest and in transit, and access control

Describe supply-chain security for AI artefacts, including model provenance, dependency and dataset integrity checks, model signing, and vetting of third-party models and plugins

Draft logging and monitoring requirements for an AI system, covering what to log, retention, tamper-evidence, and detection of anomalous inference patterns

Explain how red teaming and AI-specific penetration testing differ from conventional security testing

Distinguish AI security risk from AI governance risk, and explain where the two overlap in incident response and documentation duties

Identify the security-relevant obligations that governance frameworks place on providers and deployers, such as robustness, accuracy, logging, and human oversight

Build a risk register with at least six AI risks, each mapped to a security control and to a governance owner

Explain how technical evidence (test results, logs, model cards) supports governance claims and regulatory documentation

Explain the EU AI Act's risk-based structure: prohibited practices, high-risk systems, limited-risk transparency duties, and minimal-risk systems, and classify five example use cases correctly

Summarise the obligations on providers of high-risk systems, including risk management, data governance, technical documentation, record-keeping, transparency, human oversight, accuracy, robustness, and cybersecurity

Summarise the lighter obligations on deployers of high-risk systems, including human oversight assignment, input-data relevance, monitoring, log retention, and informing affected persons

Describe the transparency obligations for limited-risk systems, such as disclosure of AI interaction, and for synthetic content such as deepfakes and machine-readable marking

Explain the roles of notified bodies, conformity assessment, CE marking, and post-market monitoring in the compliance path

Outline the governance instruments around the Act: general-purpose AI model obligations, codes of practice, harmonised standards, and the AI literacy duty

Explain the AI Act's interaction with adjacent EU law, including GDPR, the product-safety framework, and the data-governance and platform rules, noting where duties stack

Describe prohibited practices and the rationale behind each ban, and explain how the prohibitions differ from high-risk obligations in enforcement and timing

Explain the roles and powers of market-surveillance authorities, the AI Office, and the penalty regime, including the tiered structure of fines

Draft the outline of an AI management system aligned to a recognised governance standard, covering policy, roles, risk assessment, supplier management, and continual improvement

Integrate security and governance into a single end-to-end process, from intake and classification through risk assessment, control selection, documentation, and monitoring

Produce the core documentation set for one high-risk system: technical documentation outline, risk-management record, human-oversight description, and logging plan

Design an AI incident-response playbook covering detection, triage, containment, notification duties, and post-incident review, distinguishing security incidents from serious incidents

Explain how post-market monitoring feeds back into risk management and documentation updates

Rehearse explaining a governance decision to a non-technical executive in under two minutes

Consolidate both domains into a single summary sheet of at least 40 key concepts, each defined in one sentence

Complete at least 60 practice questions under timed conditions and review every wrong answer to identify the underlying concept gap

Re-derive the AI Act risk tiers and the high-risk obligation list from memory, then check against your notes and correct errors

Re-explain the top five AI security attack patterns and their mitigations without notes

Identify your three weakest areas and write a targeted one-page revision sheet for each