Disclaimer:
Educational Use Only:
The website links provided are to be used strictly for educational and knowledge-gaining purposes only.
No Misuse:
Do not misuse the links or the information contained therein for any commercial, illegal, or unethical activities.
Respect for Websites and Businesses:
Respect the intellectual property and business interests of the websites and their owners. Do not attempt to alter, copy, or redistribute content without proper authorization.
Acknowledge Value of Knowledge:
Recognize and value the knowledge and resources shared by these websites. Use the information responsibly and ethically.
No Endorsement:
The provision of these links does not imply endorsement of the linked sites, their content, or their products and services. Users are responsible for evaluating the suitability and accuracy of the information.
User Responsibility:
By accessing these links, you agree to use the information at your own risk and to comply with all applicable laws and website terms of use.
- What is adversarial artificial intelligence in the context of cybersecurity?
a. The use of AI to generate random data for training models
b. Malicious manipulation of AI systems by feeding deceptive data
c. The process of updating machine learning models regularly
d. The use of AI only for threat detection
Answer: b. Malicious manipulation of AI systems by feeding deceptive datapaloaltonetworks - Which of the following is a primary aim of adversarial attacks in machine learning?
a. Improving prediction speed
b. Correcting biased outputs
c. Causing incorrect or unintended behavior by the AI
d. Lowering computation costs
Answer: c. Causing incorrect or unintended behavior by the AIpaloaltonetworks - Which type of attack involves corrupting the training dataset with malicious data?
a. Evasion attack
b. Poisoning attack
c. Model extraction attack
d. Membership inference attack
Answer: b. Poisoning attackpaloaltonetworks - Adversarial examples in machine learning are:
a. Randomly generated test data
b. Inputs with subtle alterations that trick models
c. Incorrect parameters fed to models during deployment
d. High-dimensional datasets
Answer: b. Inputs with subtle alterations that trick modelspaloaltonetworks - In cybersecurity, adversarial attacks can target which aspect of AI systems?
a. Only cloud storage
b. Model decision logic and data
c. Network hardware
d. Human resources
Answer: b. Model decision logic and datapaloaltonetworks - What is the most common consequence of a successful adversarial attack?
a. System upgrade
b. Model misclassification leading to faulty predictions
c. Hardware meltdown
d. Increased password strength
Answer: b. Model misclassification leading to faulty predictionspaloaltonetworks - Which adversarial attack method involves feeding crafted data to a model to mislead it after training is complete?
a. Data poisoning
b. Evasion attack
c. Model inversion
d. Password brute force
Answer: b. Evasion attackpaloaltonetworks - Which of these is NOT an adversarial attack category in machine learning?
a. Model extraction
b. Membership inference
c. Data backup
d. Poisoning attack
Answer: c. Data backuppaloaltonetworks - What differentiates adversarial attacks from traditional cyberattacks?
a. Use of older hardware
b. Exploitation of AI data processing rather than software vulnerabilities
c. Reliance on user input
d. Physical sabotage
Answer: b. Exploitation of AI data processing rather than software vulnerabilitiespaloaltonetworks - In adversarial AI, what is a “white-box” attack?
a. An attack performed without knowledge of the model
b. An attack performed with full access to the model’s architecture and parameters
c. An attack carried out by internal staff
d. An attack on encrypted data
Answer: b. An attack performed with full access to the model’s architecture and parameterspaloaltonetworks - Which method can make AI systems resilient to adversarial attacks?
a. Increasing model complexity alone
b. Adversarial training using both normal and malicious data
c. Ignoring adversarial alerts
d. Reducing input size only
Answer: b. Adversarial training using both normal and malicious datapaloaltonetworks - What are deepfake attacks mainly used for?
a. Creating realistic avatars for social engineering
b. Injecting malware into networks
c. Extracting passwords from databases
d. Optimizing code performance
Answer: a. Creating realistic avatars for social engineeringlegitsecurity - Why are adversarial attacks difficult to detect with traditional security tools?
a. They are always encrypted
b. Inputs appear legitimate while crafted to exploit model weaknesses
c. They only use outdated software
d. They always originate internally
Answer: b. Inputs appear legitimate while crafted to exploit model weaknessespaloaltonetworks - What is data poisoning in AI cybersecurity?
a. Cleaning datasets
b. Deliberately corrupting training data to produce faulty outputs
c. Enhancing model performance
d. Encrypting datasets
Answer: b. Deliberately corrupting training data to produce faulty outputslegitsecurity - Which attack involves probing a model to replicate its functionality without accessing the code?
a. Model extraction attack
b. Evasion attack
c. Poisoning attack
d. Transfer attack
Answer: a. Model extraction attackpaloaltonetworks - What is the goal of brute-force attacks with adversarial AI?
a. To guess passwords rapidly using AI
b. To analyze encrypted traffic only
c. To increase trust in AI systems
d. To limit the use of neural networks
Answer: a. To guess passwords rapidly using AIlegitsecurity - In which scenario can a minor sticker on a street sign cause a dangerous situation?
a. AI traffic management misinterprets the sign due to adversarial perturbation
b. Backup procedures fail
c. Server hardware malfunctions
d. Network switches overload
Answer: a. AI traffic management misinterprets the sign due to adversarial perturbationaryaxai - What is one major impact of successful adversarial attacks in B2B environments?
a. Faster payroll processing
b. Data breaches and loss of confidence in technology
c. Brand enhancement
d. Cheaper infrastructure
Answer: b. Data breaches and loss of confidence in technologypaloaltonetworks - Which approach helps detect adversarial inputs before reaching the model?
a. Ignoring unknown inputs
b. Input validation and transformation
c. Listing possible outputs
d. Encrypting all data
Answer: b. Input validation and transformationpaloaltonetworks - What does “membership inference attack” refer to?
a. Extracting if a specific data point was part of the training set
b. Creating new AI members for a club
c. Backing up model data to external servers
d. Increasing training set size
Answer: a. Extracting if a specific data point was part of the training setpaloaltonetworks - Which defense strategy combines multiple models for a single prediction?
a. Adversarial training
b. Ensemble methods
c. Password rotation
d. Membership inference
Answer: b. Ensemble methodspaloaltonetworks - What is the main risk of social engineering campaigns powered by AI?
a. Slow phishing attacks
b. Highly convincing fake emails, calls, or messages
c. Only visible to technical teams
d. Increased server load
Answer: b. Highly convincing fake emails, calls, or messageslegitsecurity - Why is multi-factor authentication recommended in defending against AI-powered attacks?
a. It slows down network speed
b. It increases the security layer against AI-driven brute-force password guessing
c. It makes model training easier
d. It is cheaper than passwords
Answer: b. It increases the security layer against AI-driven brute-force password guessinglegitsecurity - Which kind of adversarial input modification is especially challenging for humans to detect?
a. Obvious image distortions
b. Subtle, nearly imperceptible changes
c. Large, random alterations
d. Complete input rewriting
Answer: b. Subtle, nearly imperceptible changespaloaltonetworks - What is a “black-box” adversarial attack?
a. Attackers use random data on models
b. Attackers have limited or no knowledge of the model’s internals
c. Attackers only modify hardware
d. Attacks only work on legacy systems
Answer: b. Attackers have limited or no knowledge of the model’s internalspaloaltonetworks - Which organization found generative AI tools accelerate cybercrime entry?
a. World Economic Forum
b. MIT Technology Review
c. Crowdstrike
d. Palo Alto Networks
Answer: a. World Economic Forumlegitsecurity - What is “vishing” in the context of AI-powered attacks?
a. Video-based ransomware attack
b. Voice phishing using AI-generated voices
c. Visual hacking using deepfake images
d. Visible malware deployment
Answer: b. Voice phishing using AI-generated voiceslegitsecurity - What role do robust model architectures play in machine learning security?
a. They make retraining unnecessary
b. They increase resilience against adversarial examples
c. They slow down predictions
d. They increase network bandwidth
Answer: b. They increase resilience against adversarial examplespaloaltonetworks - What is a transfer attack in adversarial AI?
a. Reusing adversarial examples from one AI model to attack another
b. Copying training data between servers
c. Transferring predictions to human analysts
d. Shifting attack types each week
Answer: a. Reusing adversarial examples from one AI model to attack anotherpaloaltonetworks - What does erosion of trust mean in the context of AI cyberattacks?
a. People trusting AI systems more
b. Loss of confidence in organizations after AI-driven attacks
c. Restoration of damaged infrastructure
d. Trusting third-party vendors with AI data
Answer: b. Loss of confidence in organizations after AI-driven attackslegitsecurity - How can adversarial AI contribute to supply chain disruption?
a. By improving logistics
b. By targeting critical tools and services, causing outages
c. By increasing product quality
d. By enhancing communication channels
Answer: b. By targeting critical tools and services, causing outageslegitsecurity - What is the significance of anomaly detection in defending against adversarial AI?
a. Identifies normal system behavior only
b. Alerts to unusual drops in model accuracy or outputs
c. Trains models faster
d. Disables unused accounts
Answer: b. Alerts to unusual drops in model accuracy or outputspaloaltonetworks - What does the “secure development lifecycle” recommend for AI security?
a. Ignoring threat modeling
b. Integrating security from data collection through deployment
c. Only testing finished applications
d. Excluding developers from testing
Answer: b. Integrating security from data collection through deploymentpaloaltonetworks - How can regular data audits help combat data poisoning attacks?
a. By verifying and cleaning datasets for AI training
b. By speeding up predictive analytics
c. By increasing hardware capacity
d. By lowering costs
Answer: a. By verifying and cleaning datasets for AI traininglegitsecurity - Which of the following is NOT a mitigation strategy for adversarial attacks?
a. Adversarial training
b. Input validation
c. Model extraction attack
d. Ensemble methods
Answer: c. Model extraction attackpaloaltonetworks - What is “smishing” in social engineering attacks?
a. SMS phishing scams using AI
b. Smiling to gain trust
c. Cloud encryption failure
d. Server migration issues
Answer: a. SMS phishing scams using AIlegitsecurity - How were deepfake voicemails used in a cybersecurity red teaming exercise?
a. To create realistic job applications
b. To trick company employees into clicking phishing links
c. To fingerprint mobile devices
d. To increase download speeds
Answer: b. To trick company employees into clicking phishing linkslegitsecurity - Which preventative measure is effective against password hacking by AI?
a. Disabling account access
b. Multi-factor authentication and Zero Trust technology
c. Using default passwords
d. Sharing credentials
Answer: b. Multi-factor authentication and Zero Trust technologylegitsecurity - Which is the impact of open-source AI models on cybercrime?
a. Making hacking harder
b. Lowering barriers to entry for cybercriminals
c. Improving overall security
d. Limiting threats to corporations
Answer: b. Lowering barriers to entry for cybercriminalslegitsecurity - Which adversarial machine learning threat involves reconstructing sensitive data from model outputs?
a. Model inversion
b. Data backup
c. Server mirroring
d. Firmware upgrade
Answer: a. Model inversionpaloaltonetworks - What is the main risk of fake job postings in an AI-powered cyberattack?
a. Slow hiring processes
b. Obtaining personal credentials for malicious use
c. Improving remote work
d. Distributing company newsletters
Answer: b. Obtaining personal credentials for malicious uselegitsecurity - Why must organizations deploy AI-driven security platforms?
a. Manual response can’t keep pace with automated AI attacks
b. Human oversight is completely unnecessary in security
c. Such platforms are cheaper
d. AI is only for offensive security
Answer: a. Manual response can’t keep pace with automated AI attackslegitsecurity - In adversarial AI, what is the purpose of using ensemble methods?
a. To combine predictions from multiple models for resilience
b. To slow down system processing
c. To generate more adversarial examples
d. To extract model parameters
Answer: a. To combine predictions from multiple models for resiliencepaloaltonetworks - What is baiting in AI-powered social engineering?
a. Offering fake deals or lottery wins to solicit sensitive data
b. Encrypting user databases
c. Destroying malware samples
d. Speeding up transactions
Answer: a. Offering fake deals or lottery wins to solicit sensitive datalegitsecurity - What’s the easiest way for beginners to understand adversarial input perturbations?
a. Viewing perturbed images that look normal to humans but fool models
b. Watching only full system crashes
c. Ignoring image recognition examples
d. Deleting modified data
Answer: a. Viewing perturbed images that look normal to humans but fool modelspaloaltonetworks - How can employee training reduce the impact of adversarial AI threats?
a. By making staff aware of AI-driven scams and response tactics
b. By changing hardware regularly
c. By limiting access to the break room
d. By disabling user accounts
Answer: a. By making staff aware of AI-driven scams and response tacticslegitsecurity - State a defense best practice for organizations using AI.
a. Secure and validate data regularly
b. Never update software
c. Enhance only physical security
d. Limit data sharing
Answer: a. Secure and validate data regularlylegitsecurity - What do attackers exploit during the “understanding the target system” phase of adversarial AI?
a. AI model algorithms, decision-making patterns, and weaknesses
b. Only hardware specs
c. Non-essential applications
d. Spreadsheet formulas
Answer: a. AI model algorithms, decision-making patterns, and weaknessespaloaltonetworks - Which kind of attack focuses on making the AI produce any incorrect output, not necessarily a specific one?
a. Nontargeted evasion attack
b. Targeted attack
c. Backdoor attack
d. Model inversion
Answer: a. Nontargeted evasion attackpaloaltonetworks - What is the major challenge adversarial AI poses to modern authentication systems?
a. Incorrect access decisions from adversarial inputs
b. Slow login processes
c. Only increased password rotations
d. Faster account recovery
Answer: a. Incorrect access decisions from adversarial inputspaloaltonetworks
- https://www.paloaltonetworks.com/cyberpedia/what-are-adversarial-attacks-on-AI-Machine-Learning
- https://www.legitsecurity.com/aspm-knowledge-base/ai-cyberattacks
- https://www.aryaxai.com/article/understanding-adversarial-machine-learning-threats-and-challenges
- https://www.aionlinecourse.com/ai-quiz-questions/machine-learning/adversarial-learning
- https://www.gyansetu.in/blog/70-generative-ai-mcq-questions-and-answers-gyansetu/
- https://www.sanfoundry.com/1000-cyber-security-questions-answers/
- https://www.theknowledgeacademy.com/blog/cyber-security-mcq-and-answers/
- https://viso.ai/deep-learning/adversarial-machine-learning/
- https://www.javaguides.net/2023/12/generative-ai-quiz.html
- https://sysdig.com/learn-cloud-native/adversarial-ai-understanding-and-mitigating-the-threat
- https://blog.securelayer7.net/adversarial-machine-learning/
- https://www.igmguru.com/blog/cyber-security-mcqs
- https://www.infosys.com/iki/perspectives/securing-ai-adversarial-attacks.html
- https://thewhiteboard.co.in/ai-security-level-3-certification-course.php
- https://valuementor.com/blogs/adversarial-attacks
- https://en.wikipedia.org/wiki/Adversarial_machine_learning
- https://examradar.com/artificial-intelligence-ai-and-machine-learning-ml-in-cyber-security-mcqs/
- https://www.sciencedirect.com/science/article/pii/S2214212620308607
- https://mindgard.ai/blog/ai-under-attack-six-key-adversarial-attacks-and-their-consequences
- https://www.interviewbit.com/cyber-security-mcq/