CompTIA SecAI+ Certification Exam Objectives

CompTIA SecAI+ Certification Exam Objectives CY0-001 V1

About the Exam

  • The CompTIA SecAI+ certification exam validates a candidate's knowledge and skills across several critical areas related to Artificial Intelligence (AI) and cybersecurity.

  • Key Skills Certified:

    • Understanding fundamental AI concepts.

    • Securing AI systems through diverse technical controls.

    • Leveraging AI to bolster corporate security posture and automate security tasks.

    • Comprehension of how Governance, Risk, and Compliance (GRC) influence AI technologies globally.

  • Experience Equivalent: This certification is designed for professionals with an equivalent of 3−43-4 years of IT experience, including approximately 22 years of hands-on cybersecurity experience.

  • Content Examples: The provided examples serve to clarify test objectives and are not an exhaustive list of all content on the examination.

Exam Accreditation and Development

  • Accreditation: The CompTIA SecAI+ exam is accredited by the American National Standards Institute (ANSI) in adherence to the International Organization for Standardization (ISO) 1702417024 standard. This ensures regular reviews and updates to exam objectives.

  • Development: CompTIA exams are developed through subject matter expert workshops and industry-wide surveys to reflect the current skills and knowledge required of IT professionals.

CompTIA Authorized Materials Use Policy

  • CompTIA Certifications, LLC strictly prohibits the use of content from unauthorized third-party training sites (often referred to as “brain dumps”).

  • Candidates found using such materials will face revocation of their certifications and suspension from future testing, in line with the CompTIA Candidate Agreement.

  • Candidates are advised to review all CompTIA Certification Exam Policies before studying. For questions regarding unauthorized study materials, contact CompTIA at examsecurity@comptia.org.

Exam Content Updates

  • The bulleted lists of examples within the exam objectives are not exhaustive; other relevant technologies, processes, or tasks may appear on the exam.

  • CompTIA continuously reviews and updates exam content and questions to maintain currency and security.

  • Updated exams, when published, will be based on existing objectives, ensuring all related exam preparation materials remain valid.

Test Details

  • Required Exam: SecAI+ CY0-001

  • Number of questions: Undisclosed, but implied to be standard CompTIA exam quantity.

  • Types of questions: Multiple-choice and performance-based.

  • Length of test: Undisclosed.

  • Recommended experience: 3−43-4 years of IT experience and approximately 22 years of hands-on cybersecurity experience.

Exam Objectives (Domains)

  • The examination is structured around four main domains, each weighted by percentage:

    • Domain 1.0: Basic AI Concepts Related to Cybersecurity - 17 ext{%}

    • Domain 2.0: Securing AI Systems - 40 ext{%}

    • Domain 3.0: AI-assisted Security - 24 ext{%}

    • Domain 4.0: AI Governance, Risk, and Compliance - 19 ext{%}

    • Total: 100 ext{%}

1.0 Basic AI Concepts Related to Cybersecurity

1.1 Compare and contrast various AI types and techniques used in cybersecurity.

  • Types of AI:

    • Generative AI

    • Machine learning (ML)

    • Statistical learning

    • Transformers

    • Deep learning

    • Natural language processing (NLP)

      • Large language models (LLMs)

      • Small language models (SLMs)

      • Generative adversarial networks (GANs)

  • Model training techniques:

    • Model validation

    • Supervised learning

    • Unsupervised learning

    • Reinforcement learning

    • Fine-tuning

      • Epoch: A complete pass through the entire training dataset.

      • Pruning: Reducing the size of a model by removing less important connections/neurons.

      • Quantization: Reducing the precision of numbers in a model, often to 8ext−bit8 ext{-bit} integers, to decrease memory and computational requirements.

  • Prompt engineering:

    • System prompts: Instructions given to the AI system to guide its overall behavior.

    • User prompts: Input provided by the end-user to the AI.

    • One-shot prompting: Providing a single example within the prompt to guide the AI's response.

    • Multi-shot prompting: Including multiple examples in the prompt for improved guidance.

    • Zero-shot prompting: Asking the AI to perform a task without any examples.

    • System roles: Predefined roles or personas assigned to the AI system.

    • Templates: Pre-structured formats for prompts to ensure consistency and guide output.

1.2 Explain the importance of data security in relation to AI.

  • Data processing:

    • Data cleansing: Processes to detect and correct (or remove) corrupt or inaccurate records from a dataset.

    • Data verification: Ensuring the accuracy and truthfulness of data.

    • Data lineage: Tracking the origin and journey of data from source to consumption.

    • Data integrity: Maintaining and assuring the accuracy and consistency of data over its entire lifecycle.

    • Data provenance: Recording the history of data, including who accessed, modified, and used it.

    • Data augmentation: Increasing the amount of data by adding slightly modified copies of already existing data or newly created synthetic data.

    • Data balancing: Adjusting class distributions in imbalanced datasets to prevent models from being biased towards majority classes.

  • Data types:

    • Structured data: Highly organized and formatted data that resides in fixed fields within a record or file (e.g., relational databases).

    • Semi-structured data: Data that does not conform to a formal data model but has some organizational properties (e.g., JSON, XML).

    • Unstructured data: Data that does not have a predefined format or organization (e.g., text documents, images, audio, video).

  • Watermarking: Embedding covert identifiable information into data (e.g., images, text) to ascertain its origin or detect unauthorized use.

  • Retrieval-augmented generation (RAG): An AI architecture that enhances generative models by retrieving relevant information from an external knowledge base to inform its output.

    • Vector storage: Databases optimized for storing and querying high-dimensional vectors (embeddings).

    • Embeddings: Numerical representations of text or other data, capturing semantic meaning, allowing for similarity comparisons.

1.3 Explain the importance of security throughout the life cycle of AI.

  • Business use case: Ensuring the AI solution aligns with corporate objectives and ethical guidelines.

  • Data collection: Focus on trustworthiness and authenticity of data sources to prevent biased or malicious input.

  • Data preparation: Securely transforming and cleaning data, often involving anonymization and privacy controls.

  • Model development/selection: Choosing or developing models that are robust, secure, and resilient against attacks.

  • Model evaluation: Rigorous testing of the model for performance, accuracy, fairness, and security vulnerabilities.

  • Deployment: Securely integrating the AI model into production environments.

  • Validation: Continuous testing and verification of the deployed model's performance and security.

  • Monitoring and maintenance: Ongoing observation of the model's behavior, identifying anomalies, and addressing performance degradation or security issues.

  • Feedback and iteration: Incorporating user feedback and performance data to improve the model and its security over time.

  • Human-centric AI design principles:

    • Human-in-the-loop: Designing systems where humans intervene to review, validate, or override AI decisions.

    • Human oversight: Maintaining human responsibility and control over AI systems, even in autonomous operations.

    • Human validation: Ensuring human experts verify the AI's logic, outputs, and ethical adherence.

2.0 Securing AI Systems

2.1 Given a scenario, use AI threat-modeling resources.

  • Open Web Application Security Project (OWASP) Top 10:

    • LLM Top 10: Specific to vulnerabilities in Large Language Models.

    • Machine Learning (ML) Security Top 10: Focused on general ML system vulnerabilities.

  • Massachusetts Institute of Technology (MIT) AI Risk Repository: A database of risks associated with AI systems.

  • MITRE Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS): A knowledge base of adversary tactics, techniques, and common knowledge for AI systems.

  • Common Vulnerabilities and Exposures (CVE) AI Working Group: A group dedicated to identifying and standardizing AI-specific vulnerabilities.

  • Threat-modeling frameworks: Structured approaches like STRIDE, DREAD, or PASTA applied to AI systems to identify potential threats.

2.2 Given a set of requirements, implement security controls for AI systems.

  • Model controls:

    • Model evaluation: Regular assessment of model robustness, bias, and performance under adversarial conditions.

    • Model guardrails: Implementing programmatic and policy-based constraints on a model's behavior and output.

      • Prompt templates: Structured inputs that guide the model to produce desired outputs and avoid unwanted ones.

  • Gateway controls: Security measures implemented at the entry/exit points of an AI system.

    • Prompt firewalls: Filters that analyze and sanitize user prompts before they reach the AI model, blocking malicious or unsafe inputs.

    • Rate limits: Restricting the number of requests an entity can make to the AI system within a specific time frame to prevent abuse or DoS attacks.

    • Token limits: Restricting the maximum number of tokens (words/sub-words) in an AI's input or output to manage computational resources and prevent prompt injection.

    • Input quotas: Limiting the amount of data or number of inputs allowed.

      • Data size: Restricting the total volume of data.

      • Quantity: Limiting the number of inputs.

    • Modality limits: Restricting the types of media or input formats an AI can process.

    • Endpoint access controls: Securing the API endpoints through which users or other systems interact with the AI.

  • Guardrail testing and validation: Continuously testing the effectiveness of implemented guardrails against known and emerging attack vectors.

2.3 Given a scenario, implement appropriate access controls for AI systems.

  • Model access: Restricting who can view, modify, or retrain the AI model.

  • Data access: Controlling who can access the training data, inference data, and generated data.

  • Agent access: Managing permissions for AI agents or automated processes that interact with the system.

  • Network/application programming interface (API) access: Securing the network paths and API endpoints used to communicate with the AI system.

2.4 Given a scenario, implement data security controls for AI systems.

  • Encryption requirements: Protecting data confidentiality.

    • In transit: Encrypting data as it moves across networks.

    • At rest: Encrypting data stored on disks or in databases.

    • In use: Protecting data while it is being processed in memory (e.g., homomorphic encryption, confidential computing).

  • Data safety: Techniques to minimize risks associated with sensitive data.

    • Data anonymization: Removing or encrypting personally identifiable information (PII) from datasets.

    • Data classification labels: Tagging data based on its sensitivity to enforce appropriate handling policies.

    • Data redaction: Permanently removing sensitive information from a document or dataset.

    • Data masking: Obscuring sensitive data with realistic but false information for testing or development purposes.

    • Data minimization: Collecting and processing only the absolutely necessary data required for a specific purpose.

2.5 Given a scenario, implement monitoring and auditing for AI systems.

  • Prompt monitoring:

    • Query: Analyzing user input prompts for malicious intent, sensitive information, or policy violations.

    • Response: Examining the AI's generated outputs for undesirable content, hallucinations, or data leakage.

  • Log monitoring: Collecting and analyzing system logs for security events, performance issues, and anomalies.

  • Log sanitization: Removing sensitive or personally identifiable information from logs before storage or analysis.

  • Log protection: Securing log files from unauthorized access or tampering.

  • Response confidence level: Tracking the AI model's certainty in its predictions or generated output to identify potential errors or weaknesses.

  • Rate monitoring: Tracking the frequency of requests or operations to detect Denial of Service (DoS) attacks or misuse.

  • AI cost monitoring: Tracking resource consumption associated with AI operations.

    • Prompts: Cost associated with processing user inputs.

    • Storage: Cost of storing datasets, models, and outputs.

    • Response: Cost associated with generating AI outputs.

    • Processing: Computational cost of model training and inference.

  • Auditing for quality and compliance: Regularly assessing the AI system against predefined standards.

    • Hallucinations: Detecting instances where the AI generates plausible but false information.

    • Accuracy: Verifying the correctness of the AI's predictions or outputs.

    • Bias and fairness: Testing for and mitigating unfair or discriminatory outcomes.

    • Access: Auditing who accessed the AI system, when, and for what purpose.

2.6 Given a scenario, analyze the evidence of an attack and suggest compensating controls for AI systems.

  • Attacks:

    • Prompt injection: Crafting malicious inputs to override or manipulate the AI's intended behavior.

    • Poisoning:

      • Model poisoning: Introducing malicious data during training to compromise the model's integrity or performance.

      • Data poisoning: Injecting corrupted or misleading data into the training dataset.

    • Jailbreaking: Bypassing an AI's safety filters or ethical guardrails to elicit prohibited responses.

    • Hallucinations: AI generating confident but incorrect or fabricated information.

    • Input manipulation: Tampering with input data to deceive the model.

    • Introducing biases: Deliberately feeding biased data into a model to produce unfair outcomes.

    • Circumventing AI guardrails: Finding ways around the protective measures implemented in an AI system.

    • Manipulating application integrations: Exploiting vulnerabilities in how an AI system connects with other applications.

    • Model inversion: Reconstructing sensitive training data from the model's outputs.

    • Model theft: Illegally acquiring or copying an AI model.

    • AI supply chain attacks: Targeting any stage of the AI development and deployment pipeline (e.g., data, libraries, infrastructure).

    • Transfer learning attacks: Exploiting models trained on one task and fine-tuned for another.

    • Model skewing: Causing the model to drift in its predictions over time due to manipulated data or environment.

    • Output integrity attacks: Tampering with the AI's output after generation but before it reaches the user.

    • Membership inference: Determining if a specific data record was part of the model's training dataset.

    • Insecure output handling: Improper storage or display of AI outputs, leading to vulnerabilities.

    • Model denial of service (DoS): Overwhelming an AI model with requests or complex inputs to reduce its availability or performance.

    • Sensitive information disclosure: AI inadvertently revealing confidential data.

    • Insecure plug-in design: Vulnerabilities arising from poorly designed or implemented AI plug-ins.

    • Excessive agency: AI agents performing actions beyond their intended scope or authority.

    • Overreliance: Humans excessively trusting or depending on AI outputs without critical judgment, leading to errors.

  • Compensating controls:

    • Prompt firewalls: Filtering and validating prompt inputs.

    • Model guardrails: Implementing behavioral constraints and safety policies.

    • Access controls: Limiting who can interact with the AI model and its data.

    • Data integrity controls: Ensuring the accuracy and consistency of data throughout its lifecycle.

    • Encryption: Protecting data in all states (at rest, in transit, in use).

    • Prompt templates: Guiding valid inputs and preventing malicious ones.

    • Rate limiting: Controlling the frequency of interactions to prevent abuse.

    • Least privilege: Granting only the minimum necessary permissions to users and systems.

3.0 AI-assisted Security

3.1 Given a scenario, use AI-enabled tools to facilitate security tasks.

  • Tools/applications:

    • Integrated development environment (IDE) plug-ins: AI tools integrated into coding environments for security analysis.

    • Browser plug-ins: AI-powered tools within web browsers for security-related tasks.

    • Command-line interface (CLI) plug-ins: AI tools accessible and operable from the command line.

    • Chatbots: Conversational AI agents assisting with security queries, initial incident response, or information retrieval.

    • Personal assistants: AI-powered digital assistants for security-related tasks, scheduling, or alerts.

  • Use cases:

    • Signature matching: Automated detection of known malware signatures.

    • Code quality and linting: AI-assisted analysis of code for errors, style violations, and potential vulnerabilities.

    • Vulnerability analysis: Identifying security weaknesses in systems and applications.

    • Automated penetration testing: AI-driven simulations of attacks to find exploitable vulnerabilities.

    • Anomaly detection: Identifying unusual patterns that may indicate a security incident.

    • Pattern recognition: Discovering recurring behaviors or data sequences indicative of threats.

    • Incident management: Automating parts of the incident response lifecycle, such as triage and escalation.

    • Threat modeling: Assisting in identifying potential threats and vulnerabilities in systems.

    • Fraud detection: Identifying fraudulent transactions or activities using behavioral analysis.

    • Translation: Translating security intelligence or incident reports across languages.

    • Summarization: Condensing large security logs or reports into concise summaries.

3.2 Explain how AI enables or enhances attack vectors.

  • AI-generated content (deepfake):

    • Impersonation: Creating realistic fake audio/video to impersonate individuals for scams or social engineering.

    • Misinformation: Spreading inaccurate or misleading information.

    • Disinformation: Deliberately spreading false information to deceive.

  • Adversarial networks: Using GANs or similar models to generate adversarial examples or learn optimal attack strategies.

  • Reconnaissance: AI automating collection and analysis of target information (e.g., open-source intelligence).

  • Social engineering: AI crafting highly convincing phishing emails, voice messages, or social media interactions.

  • Obfuscation: AI generating complex and evasive code or techniques to hide malicious activity from security tools.

  • Automated data correlation: AI systems quickly identifying relationships and patterns across vast datasets to find vulnerabilities or attack pathways.

  • Automated attack generation: AI creating sophisticated and novel attacks.

    • Attack vector discovery: Identifying new ways to exploit systems.

    • Payloads: Generating custom malware payloads.

    • Malware: Creating polymorphic or metamorphic malware to evade detection.

    • Honeypot: AI designing and managing deception environments to trap attackers or gather intelligence.

    • Distributed denial of service (DDoS): Orchestrating and coordinating large-scale attacks from distributed sources.

3.3 Given a scenario, use AI to automate security tasks.

  • Scripting tools:

    • Low-code: Tools allowing developers to create applications with minimal manual coding, often using visual interfaces, enabling faster security automation.

    • No-code: Tools allowing non-developers to create applications with no coding, empowering broader security automation.

  • Document synthesis and summarization: AI-powered tools to automatically generate security reports or summarize incident logs.

  • Incident response ticket management: AI automating the creation, categorization, assignment, and tracking of security incident tickets.

  • Change management:

    • AI-assisted approvals: AI reviewing change requests for potential security risks and recommending approvals or rejections.

    • Automated deployment/rollback: AI managing the automated deployment of security patches or configuration changes and enabling quick rollbacks in case of issues.

  • AI agents: Autonomous AI entities performing security tasks like monitoring, threat hunting, or vulnerability remediation.

  • Continuous integration/continuous deployment (CI/CD): Integrating AI into the software development pipeline to enhance security automation.

    • Code scanning: Automated analysis of source code for security vulnerabilities.

    • Software composition analysis (SCA): Identifying known vulnerabilities in open-source and third-party components.

    • Unit testing: Automating tests for individual code components.

    • Regression testing: Ensuring new code changes don't break existing functionalities or introduce new vulnerabilities.

    • Model testing: Automated testing of AI models for biases, robustness, and performance.

    • Automated deployment/rollback: AI managing the automated deployment of applications and ensuring secure rollbacks if problems arise.

4.0 AI Governance, Risk, and Compliance

4.1 Explain organizational governance structures that support AI.

  • Organizational structures:

    • AI Center of Excellence: A dedicated team or department focused on guiding AI strategy, best practices, and governance across an organization.

    • AI policies and procedures: Formal documents outlining rules, guidelines, and processes for the responsible and secure use of AI.

  • AI-related roles:

    • Data scientist: Designs and implements AI models, focusing on data analysis and model development.

    • AI architect: Designs the overall architecture and infrastructure for AI systems.

    • Machine learning engineer: Builds, deploys, and maintains ML models in production.

    • Platform engineer: Develops and maintains the underlying infrastructure and platforms that support AI development and deployment.

    • MLOps engineer: Specializes in operationalizing ML models, focusing on CI/CD, monitoring, and scaling of AI systems.

    • AI security architect: Designs and implements security measures for AI systems and infrastructure.

    • AI governance engineer: Ensures AI systems comply with ethical guidelines, regulations, and internal policies.

    • AI risk analyst: Identifies, assesses, and mitigates risks associated with AI deployment.

    • AI auditor: Evaluates AI systems for compliance, fairness, and performance against established standards.

    • Data engineer: Develops and maintains data pipelines and data infrastructure to ensure data availability and quality for AI applications.

4.2 Explain risks associated with AI.

  • Responsible AI: A framework emphasizing ethical and societal considerations in AI development and deployment.

    • Fairness: Ensuring AI systems do not produce discriminatory or biased outcomes.

    • Reliability and safety: Ensuring AI systems perform consistently and safely without unintended harm.

    • Transparency: Making AI decision-making processes understandable and explainable.

    • Privacy and security: Protecting sensitive data and securing AI systems from attacks.

    • Explainability: The ability to describe why an AI system made a particular decision or prediction.

    • Inclusiveness: Designing AI systems to be accessible and beneficial to all users, regardless of background.

    • Accountability: Establishing clear responsibility for the actions and impacts of AI systems.

    • Consistency: Ensuring AI systems behave predictably and do not deviate arbitrarily.
      R* Risks:

    • Introduction of bias: AI models reflecting and amplifying biases present in training data.

    • Accidental data leakage: Inadvertent exposure of sensitive data by AI systems.

    • Reputational loss: Damage to an organization's public image due to AI's unethical behavior or security failures.

    • Accuracy and performance of the model: Issues leading to incorrect decisions or failures, potentially with severe consequences.

    • Intellectual Property (IP)-related risks: AI systems potentially infringing on existing IP or having their own IP stolen.

    • Autonomous systems: Risks associated with AI systems operating independently, including unexpected behavior or loss of human control.

4.3 Summarize the impact of compliance on business use and development of AI.

  • European Union (EU) AI Act: Landmark regulation aiming to ensure AI systems used in the EU are safe, transparent, non-discriminatory, and environmentally friendly.

  • Organisation for Economic Co-operation and Development (OECD) standards: International principles and recommendations for responsible AI innovation and stewardship.

  • ISO AI standards: International Organization for Standardization documents providing guidance and frameworks for AI systems, including quality, security, and ethics.

  • National Institute of Standards and Technology (NIST) AI Risk Management Framework (AI RMF): A voluntary framework that helps organizations manage risks associated with AI systems.

  • Corporate policies: Internal guidelines dictating AI usage.

    • Sanctioned vs. unsanctioned: Differentiating between officially approved and unapproved AI tools/practices.

    • Private vs. public models: Guidelines for using internal, proprietary models versus publicly available ones.

    • Sensitive data governance: Policies for handling, storing, and processing sensitive data with AI.

  • Third-party compliance evaluations: Assessments by external auditors to ensure an organization's AI practices meet regulatory and ethical standards.

CompTIA SecAI+ Acronym List

  • AI: Artificial Intelligence

  • API: Application Programming Interface

  • ATLAS: Adversarial Threat Landscape for Artificial-Intelligence Systems

  • CI/CD: Continuous Integration/Continuous Deployment

  • CLI: Command-line Interface

  • CVE: Common Vulnerabilities and Exposures

  • DDoS: Distributed Denial of Service

  • DoS: Denial of Service

  • EU: European Union

  • GAN: Generative Adversarial Network

  • GPU: Graphics Processing Unit

  • GRC: Governance, Risk, and Compliance

  • IDE: Integrated Development Environment

  • IP: Intellectual Property

  • ISO: International Organization for Standardization

  • LAN: Local Area Network

  • LLM: Large Language Model

  • MIT: Massachusetts Institute of Technology

  • ML: Machine Learning

  • NIST: National Institute of Standards and Technology

  • NLP: Natural Language Processing

  • OECD: Organisation for Economic Co-operation and Development

  • OWASP: Open Web Application Security Project

  • RAG: Retrieval-augmented Generation

  • RMF: Risk Management Framework

  • SCA: Software Composition Analysis

  • SLM: Small Language Model

CompTIA SecAI+ Hardware and Software List (Sample)

  • Equipment:

    • Laptops

    • Cloud VMs (Virtual Machines)

    • Graphics processing units (GPUs)

    • NVidia Jetson Nano Orin

    • Mobile devices

    • Sandbox environment

    • Local area network (LAN)

  • Software:

    • Virtual containers

    • Large data sets

    • Test data sets

    • Python environment

    • R environment

    • IDE (Integrated Development Environment)

    • Jupyter environment

    • Chatbots

    • LLMs (Large Language Models)

    • Open-source tools:

      • GitHub

      • Ollama

    • Cloud-based environment

    • Cloud-based AI studios

    • Vector database

    • NoSQL Database

    • Neo4j Graph Database