Perturbation methods

biblio

Evading Deepfake-Image Detectors with White- and Black-Box Attacks

Submitted by grigby1 on Thu, 03/04/2021 - 1:35pm

White Box Security
white box
security
attack case studies
AUC
black-box attack
composability
deepfake-image detectors
disinformation campaigns
Forensics
fraudulent social media profiles
Generators
image area
image classification
Image forensics
image generators
image representation
image sensors
image-forensic classifiers
learning (artificial intelligence)
Metrics
neural nets
neural network
optimization
Perturbation methods
popular forensic approach
pubcrawl
resilience
Resiliency
Robustness
security of data
significant vulnerabilities
social networking (online)
state- of-the-art classifier
synthesizer
synthetic content
synthetically-generated content
target classifier
Training
Twitter

biblio

Defending Against Model Stealing Attacks With Adaptive Misinformation

Submitted by grigby1 on Thu, 01/28/2021 - 12:12pm

Adaptation models
Adaptive Misinformation
Adversary Models
attacker
attacker clone model
black-box query access
clone model
Cloning
Computational modeling
Data models
deep neural networks
Human behavior
labeled dataset
learning (artificial intelligence)
Metrics
model stealing attacks
neural nets
OOD queries
out-of-distribution inputs
Perturbation methods
Predictive models
pubcrawl
query processing
resilience
Resiliency
Scalability
security
security of data
training dataset

biblio

Enhancing Resilience of Deep Learning Networks By Means of Transferable Adversaries

Submitted by grigby1 on Thu, 01/28/2021 - 12:10pm

Adversarial training
Adversary Models
Artificial Neural Networks
classification decision model
deep learning
deep learning networks
defense methods
Human behavior
Information systems
learning (artificial intelligence)
machine learning
Metrics
Multi-step Adversaries
neural nets
Neural networks
pattern classification
Perturbation methods
pubcrawl
resilience
resilience enhancement
Resiliency
Scalability
security of data
single-step adversaries
Task Analysis
Training
transferable adversaries

biblio

Link Prediction Adversarial Attack Via Iterative Gradient Attack

Submitted by aekwall on Mon, 01/25/2021 - 12:01pm

security of data
data privacy
Task Analysis
learning (artificial intelligence)
Resiliency
pubcrawl
composability
graph theory
privacy
Robustness
neural nets
Prediction algorithms
security problem
Predictive models
Predictive Metrics
deep neural networks
Perturbation methods
gradient methods
attack graphs
adversarial attack
adversarial graph
deep models
defense
GAE
gradient attack (GA)
gradient attack strategy
gradient information
graph autoencode
graph evolved tasks
iterative gradient attack
Link prediction
link prediction adversarial attack problem
node classification
real-world graphs
trained graph autoencoder model

biblio

Adversarial Perturbations Fool Deepfake Detectors

Submitted by grigby1 on Fri, 01/15/2021 - 11:02am

Training
Unsupervised Learning
convolutional neural nets
Neural networks
pubcrawl
Human behavior
Human Factors
Metrics
Resiliency
Scalability
Perturbation methods
Electronics packaging
resilience
gradient methods
adversarial perturbations
feature extraction
Detectors
generative adversarial networks
image enhancement
DeepFake
faces
Deep Image Prior
deepfake detectors
deepfake image enhancement
Deepfakes
fast gradient sign method
generative convolutional neural networks
image restoration
Lipschitz regularization

biblio

An Improved Collaborative Filtering Recommendation Algorithm Based on Differential Privacy

Submitted by aekwall on Mon, 01/11/2021 - 12:40pm

Scalability
data privacy
Resiliency
Human behavior
pubcrawl
composability
collaborative filtering
privacy
Prediction algorithms
Stochastic processes
software engineering
Predictive models
differential privacy
privacy protection
recommender systems
Perturbation methods
gradient methods
differential privacy protection
Matrix decomposition
matrix factorization
centralized recommendation
collaborative filtering recommendation model
differential privacy constraint
differential privacy matrix factorization model
MovieLens
Netflix
noise matrix factorization model
potential characteristic matrix
recommendation accuracy loss
recommendation systems
stochastic gradient descent algorithm
ε-differential privacy

biblio

BlurNet: Defense by Filtering the Feature Maps

Submitted by grigby1 on Mon, 12/28/2020 - 11:48am

Adaptation models
adaptive attack evaluation
adaptive filtering
adversarial defense
adversarial images
Adversarial Machine Learning
Adversarial robustness
attack algorithms
black stickers
blackbox transfer attack
BlurNet
depthwise convolution layer
frequency analysis
gradient information
high frequency noise
image recognition
image restoration
input image
Kernel
layer feature maps
learning (artificial intelligence)
low-pass filters
lowpass filtering behavior
malicious adversary
malicious examples
Metrics
neural nets
Neural networks
Perturbation methods
pubcrawl
resilience
Resiliency
robust physical perturbations
Robustness
RP
Scalability
security of data
standard blur kernels
standard-architecture traffic sign classifiers
standards
stop signs
substitute model
targeted misclassification rates
traffic engineering computing
victim model
white stickers
white-box attacks

biblio

A Study of Big Data Security on a Partitional Clustering Algorithm with Perturbation Technique

Submitted by grigby1 on Mon, 12/28/2020 - 11:40am

Accuracy
Big Data
big data privacy
big data security
Clustering algorithms
Clustering Tme
Conferences
data privacy
Distributed databases
execution time
F-Socre Measure
Hadoop distributed file system
HDFS
Human behavior
Metrics
parallel processing
partitional clustering algorithm
Partitioning algorithms
pattern clustering
PCA
Perturbation methods
perturbation technique
Precision
principal component analysis
privacy preserving
pubcrawl
resilience
Resiliency
Scalability
security
security of data

biblio

Real-Time Attack Detection on Robot Cameras: A Self-Driving Car Application

Submitted by grigby1 on Thu, 12/17/2020 - 12:37pm

Attack detection
automotive electronics
Cameras
car application
Human behavior
Human Factors
image matching
images comparisons
images processing
intrusion detection system
mobile robots
object detection
operating systems
Perturbation methods
policy-based governance
pubcrawl
real-time attack detection
resilience
Resiliency
Robot
robot cameras
Robot Operating System
robot operating systems
Robot vision systems
ROS
security
security assessment
security of data
self-driving cars
Service robots
telecommunication security
unsupervised anomaly detection method

biblio

Fidelity: Towards Measuring the Trustworthiness of Neural Network Classification

Submitted by aekwall on Mon, 12/07/2020 - 11:32am

security of data
Task Analysis
Trusted Computing
learning (artificial intelligence)
pubcrawl
composability
Computational modeling
Sociology
Statistics
neural nets
Neural networks
pattern classification
machine learning
machine learning model
trustworthiness
Perturbation methods
adversarial examples
adversarial attack detection
adversarial settings
neural network classification
neural network system
security-critical tasks

1
2
3
4
5
next ›
last »

Cyber-Physical Systems Virtual Organization

Fostering collaboration among CPS professionals in academia, government, and industry

Perturbation methods

Evading Deepfake-Image Detectors with White- and Black-Box Attacks

Defending Against Model Stealing Attacks With Adaptive Misinformation

Enhancing Resilience of Deep Learning Networks By Means of Transferable Adversaries

Link Prediction Adversarial Attack Via Iterative Gradient Attack

Adversarial Perturbations Fool Deepfake Detectors

An Improved Collaborative Filtering Recommendation Algorithm Based on Differential Privacy

BlurNet: Defense by Filtering the Feature Maps

A Study of Big Data Security on a Partitional Clustering Algorithm with Perturbation Technique

Real-Time Attack Detection on Robot Cameras: A Self-Driving Car Application

Fidelity: Towards Measuring the Trustworthiness of Neural Network Classification