MCQs

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.

Overall MCQs for Reference

What is AI-based steganography primarily used for?
a. Training neural networks
b. Concealing information inside digital media
c. Speeding up image compression
d. Improving cloud data access
Answer: bscribd+1
Which component of an image is most commonly manipulated to hide data using steganography?
a. Brightness
b. Saturation
c. Least Significant Bit (LSB)
d. Most Significant Bit (MSB)
Answer: cscribd
What is the main objective of data hiding in Artificial Intelligence applications?
a. Enhancing user experience
b. Making content undetectable
c. Image segmentation
d. Data compression
Answer: bwayground+1
In the context of adversarial machine learning, what can “hiding” refer to?
a. Compressing training data
b. Masking the presence or effects of adversarial modifications
c. Improving prediction speed
d. Encrypting user passwords
Answer: bhiddenlayer+1
Which file format is commonly chosen for steganography due to its uncompressed pixel values?
a. JPEG
b. PNG
c. BMP
d. GIF
Answer: cscribd
Which method can AI employ to evade detection of hidden attacks in image classifiers?
a. Overfitting the training set
b. Manipulating visual explanations such as Grad-CAM
c. Removing hidden layers from the neural network
d. Reducing batch size
Answer: binovex
What is the process of analyzing an AI-generated media file to detect hidden content called?
a. Data mining
b. Steganalysis
c. File scanning
d. Cryptanalysis
Answer: bwayground
Which term best describes embedding secret messages inside ordinary communication for covert purposes?
a. Steganography
b. Encryption only
c. Decompression
d. Compression
Answer: awayground+1
What can adversarial attacks in AI aim to hide from detection?
a. Image pixel values
b. Malicious intent and true data manipulations
c. Hardware malfunctions
d. Network latency
Answer: barxiv+1
Which adversarial ML technique involves subtly modifying input data to fool models?
a. Data quantization
b. Adversarial example generation
c. Data anonymization
d. Streamlining computation
Answer: bdatacamp+1
Which layer in a neural network is often associated with hidden processing?
a. Input layer
b. Output layer
c. Processing layer
d. Hidden layer
Answer: dcsiplearninghub
What does “cover image” refer to in digital steganography?
a. An image used as a template for detection
b. The original image used to embed secret information
c. The encrypted output
d. The color palette
Answer: bscribd
What makes steganography detection challenging for AI systems?
a. It is always compressed
b. High false positive rates
c. Does not support encryption
d. Memory leaks
Answer: bwayground
What is the primary purpose of steganalysis in AI security research?
a. To compress audio files
b. To detect the presence and location of hidden data in media
c. To accelerate training
d. To encrypt content
Answer: bwayground+1
Which part of a convolutional neural network (CNN) can be manipulated to hide adversarial intent?
a. Convolutional kernel size
b. Feature maps and their explanations
c. Data labels
d. Training epochs
Answer: binovex
What is the goal of an intelligently hidden adversarial attack?
a. Maximizing computation cost
b. Concealing malicious modifications so outputs appear normal
c. Increasing image quality
d. Deleting model parameters
Answer: barxiv+1
A typical feature of machine learning models attacked through hiding techniques is:
a. Enhanced generalization
b. Undetected malicious alterations
c. Fast training convergence
d. Increased model size
Answer: bhiddenlayer+1
What is the process of deliberately hiding the real intention of a user prompt from an AI model called?
a. Data augmentation
b. Adversarial question decomposition
c. Overfitting
d. Vectorization
Answer: barxiv
Which AI defense involves modifying overt data to disrupt hidden messages?
a. Decoding
b. Erasing
c. Detection
d. Encoding
Answer: bwayground
What is neural network watermarking often used for?
a. Hiding authorship or ownership information inside ML models
b. Compressing weights
c. Avoiding overfitting
d. Enhancing model speed
Answer: awayground
Which of these can serve as a container for steganographic hiding in AI?
a. Audio files
b. Video files
c. Digital images
d. All of the above
Answer: dstudocu+1
How does least significant bit (LSB) steganography work?
a. Uses most significant pixel values
b. Replaces the smallest bits in image data with secret information
c. Encrypts each color channel
d. Inverts the image
Answer: bscribd
Which process involves repeatedly modifying model explanations to avoid exposing manipulations?
a. Data compression
b. Adversarial fine-tuning
c. Quantization
d. Overfitting
Answer: binovex
What kind of ML attack involves making input alterations invisible to the human eye, but fooling the model?
a. Poisoning attack
b. Backdoor attack
c. Adversarial example
d. Model extraction
Answer: cdatacamp+1
Why are BMP images often a target for hiding in AI steganography?
a. They have a fixed size
b. They are uncompressed and pixel modifications are less noticeable
c. Easy to convert to PDF
d. They use CMYK color
Answer: bscribd
Which of the following is a key objective in hiding adversarial attacks on ML models?
a. Retain correct classification on clean data
b. Ensure attacks are misclassified
c. Manipulate explanations to hide modifications
d. All of the above
Answer: dinovex
What does an intelligent adversary often try to hide from an AI model?
a. Their geographical location
b. The true intention or outcome of the query
c. Training epoch number
d. Output file format
Answer: barxiv
In audio steganography, data is often hidden by:
a. Inverting waveforms
b. Altering imperceptible audio signals
c. Using mono instead of stereo
d. Speeding up playback
Answer: bscribd
What is a likely defense to hidden adversarial attacks in AI?
a. Ignoring input sanitization
b. Enhancing model explainability and analyzing explanations
c. Increasing batch size
d. Reducing computation
Answer: bdatacamp+1
Which of the following is NOT an application of steganography in AI and digital security?
a. Hiding copyright markers
b. Embedding malware in images
c. Accelerating GPU rendering
d. Covert communication
Answer: cwayground
Which technique can be used to generate adversarial examples for hiding attacks?
a. Fast Gradient Sign Method (FGSM)
b. Max Pooling
c. Cross-entropy loss
d. Dropout regularization
Answer: ahiddenlayer
Which color component is typically less noticeable when modified for hiding content?
a. Red
b. Blue
c. Green
d. Alpha
Answer: bscribd
Digital watermarking in AI is designed to:
a. Hide a signature or identifier in data
b. Encrypt text files
c. Speed up computation
d. Filter stopwords
Answer: awayground
What does the practice of “erasing” mean in steganalysis?
a. Modifying data to remove possible hidden content
b. Deleting image metadata
c. Shrinking image size
d. Encrypting content
Answer: awayground
Hidden layer manipulation in neural networks usually affects:
a. Output clarity
b. Information flow between input and output
c. File encryption
d. Model explainability only
Answer: bcsiplearninghub
Steganography combined with adversarial AI may be used for:
a. Automatic bug fixing
b. Concealing attack code in benign-looking data
c. Data normalization
d. Hyperparameter tuning
Answer: bhiddenlayer+1
What is a commonly cited limitation of steganography detection tools in AI?
a. Cannot process PNGs
b. High rate of false positives
c. Only works on text files
d. Too slow for images
Answer: bwayground
Hiding adversarial attacks via explainable AI typically involves:
a. Creating explanations that look legitimate for both clean and manipulated data
b. Removing output layer
c. Model distillation
d. Compressing data
Answer: ainovex
Which is NOT a typical container for steganographic data hiding?
a. Image
b. Video
c. Audio
d. Database schema
Answer: dscribd
Adversarial ML attacks that are “hidden” are generally:
a. Easily visible to human analysts
b. Subtle and hard to detect without advanced methods
c. Illegal in all contexts
d. Cloud-based only
Answer: bdatacamp+1
Least significant bit techniques often fail when:
a. Files are compressed after hiding
b. Original image is grayscale
c. File is opened in Photoshop
d. File format is PDF
Answer: ascribd
Which AI component can be used as a covert communication channel?
a. Model weights
b. Hidden layers
c. Bias values
d. Any of the above
Answer: dwayground+1
A main motivation for hiding information in AI models is:
a. Unauthorized access
b. Protecting intellectual property and ensuring covert transmission
c. Compressing files
d. Increasing model speed
Answer: bwayground+1
One method attackers use to avoid detection in adversarial attacks is:
a. Splitting malicious queries into benign sub-questions
b. Increasing model capacity
c. Randomizing weights
d. Using larger datasets
Answer: aarxiv
What makes feature map manipulation risky in adversarial hiding?
a. Lost color depth
b. If detected, the attack is easily attributed to the source
c. Increased computational load
d. Slower training times
Answer: binovex
Why is explainable AI important for security against hidden attacks?
a. It speeds up training
b. Allows analysis of both model decisions and potential manipulations
c. Reduces required computation
d. Prevents all hacks
Answer: bdatacamp+1
In the context of hiding, what is a “payload”?
a. The host file
b. The secret information being hidden
c. The compression algorithm
d. The network protocol
Answer: bstudyguides+1
How can adversarial examples be further hidden from model auditors?
a. Manipulating outputs only
b. Also disguising explanation maps to look similar to clean data
c. Lowering model capacity
d. Using only sequential data
Answer: binovex
Which ML vulnerability exploits data hiding to bypass AI security?
a. Model stealing
b. Adversarial steganography
c. Feature selection
d. Data quantization
Answer: bdatacamp+1
Which statement is true about hiding techniques in AI?
a. They are foolproof and undetectable
b. They can often be detected by explainable AI and steganalysis, but sophisticated variants are difficult to spot
c. Only used for legal data hiding
d. Require high computational resources only
Answer: bwayground+2

Blog

Systemethic Technologies

Leave a Reply Cancel reply

Useful links

Other Terms & Policies

Contact Info