Adversarial policies beat superhuman Go AI systems.
problem Vulnerability of superhuman AI systems to adversarial attacks.
method Training adversarial policies to trick KataGo into making blunders.
result Adversarial policies achieve >97% win rate against KataGo at superhuman settings.
Paper examines Go AI robustness against adversarial attacks.
problem Superhuman Go AIs are vulnerable to simple adversarial strategies.
method Three defenses tested: adversarial training, iterated adversarial training, and changing network architecture.
result No defense is robust against newly trained adversaries, and attacks are similar to cyclic attacks.
AI-GAN generates realistic adversarial examples efficiently.
problem Efficient generation of perceptually realistic adversarial examples.
method A joint training of a generator, discriminator, and attacker.
result Achieves high attack success rates and reduces generation time.
AI attacks threaten insurance systems, requiring new defenses.
problem Adversarial attacks on AI in insurance.
method Categorize and discuss various types of attacks and defense methods.
result Need for improved AI systems to resist attacks.
Article evaluates AI security threats and proposes multiple measures.
problem Threats to AI integrity and security.
method Literature review, analysis of AI supply chain, discussion of mitigations.
result Multiple protective measures are necessary for AI security.
Quantile regression attacks outperform shadow models in unseen class membership inference attacks.
problem Failure of shadow model attacks on unseen classes due to limited data access.
method Quantile regression attacks that learn features of member examples.
result Quantile regression attacks achieve up to 11x higher TPR than shadow model-based approaches.
Low-rate application layer distributed denial of service (LDDoS) attacks are both powerful and stealthy. They force vulnerable webservers to open all available connections to the adversary, denying resources to real users. Mitigation advice focuses on solutions that potentially degrade quality of service for legitimate…
New attacks improve privacy audits by analyzing model updates.
problem Improving privacy audits of AI models through sequence analysis.
method Developed SeMI attacks to identify target insertions in model sequences.
result SeMI attacks achieve higher power and tighter privacy audits.
This paper improves AI defenses against network attacks using ML and adversarial learning.
problem Protecting personal data from sophisticated network attacks.
method Unified multi-modal dataset, machine learning for detection, adversarial learning for synthetic data generation.
result Stable ML models for intrusion detection and high-fidelity synthetic data.
New attack targets speech-based AI models via stock market data.
problem Vulnerabilities in speech-based AI models using LLMs.
method Backdoor attack using acoustic data poisoning.
result Shows possible vulnerabilities in speech-based transformers.
CAVs reveal latent concept distributions, but are vulnerable to adversarial attacks.
problem Understanding latent concept encodings in AI models.
method Probabilistic perspective on CAVs, deriving mean and covariance.
result CAVs can be adversarially manipulated, highlighting a vulnerability.
DeFi TrustBoost uses blockchain and AI to assess small business loans.
problem Assessing small business loans from low-wealth households.
method Combines blockchain and Explainable AI to ensure confidentiality, compliance, and security.
result Tamper-proof auditing and on-chain/off-chain data storage for financial organizations.
BadGD identifies gradient descent vulnerabilities through strategic backdoor attacks.
problem Gradient descent vulnerabilities through malicious data manipulation.
method Introduces Max RiskWarp, Max GradWarp, and Max GradDistWarp triggers to exploit gradient descent.
result Demonstrates how malicious triggers can significantly alter loss landscapes and gradient calculations.
Insurance contracts for autonomous AI agents must be actuarially sound and resistant to gaming.
problem Designing insurance contracts for autonomous AI agents that are actuarially sound and resistant to gaming.
method Characterizing a five-attack space and proving the actuarial runtime is gaming-resistant.
result An incentive-compatible layer for actuarial control of autonomous-agent side effects.
Deep-learning based classification algorithms have been shown to be susceptible to adversarial attacks: minor changes to the input of classifiers can dramatically change their outputs, while being imperceptible to humans. In this paper, we present a simple hypothesis about a feature compression property of artificial i…
1-Lipschitz neural networks produce clearer, more focused Saliency Maps for explainable AI.
problem Noisy and limited Saliency Maps from traditional neural networks.
method Dual loss of optimal transport problem for 1-Lipschitz neural networks.
result Saliency Maps from 1-Lipschitz networks are highly concentrated and less noisy, aligning with human explanations.
The success of modern Artificial Intelligence (AI) technologies depends critically on the ability to learn non-linear functional dependencies from large, high dimensional data sets. Despite recent high-profile successes, empirical evidence indicates that the high predictive performance is often paired with low robustne…
FDA3 defends IIoT applications against adversarial attacks by federating defense knowledge.
problem Adversarial attacks on DNNs in IIoT applications can cause devastating consequences.
method Federated learning approach to aggregate defense knowledge from different sources.
result FDA3 can resist more attacks than existing methods and prevent new attacks.
AI models aligned with human vision perform well on few data tasks.
problem Few-shot learning performance with limited data.
method Information-theoretic analysis and empirical testing of 491 models.
result Highly aligned models show better robustness to attacks and domain shifts.
Establishing unique identities for both humans and end systems has been an active research problem in the security community, giving rise to innovative machine learning-based authentication techniques. Although such techniques offer an automated method to establish identity, they have not been vetted against sophistica…
Survey of AI-based outlier detection methods for various domains.
problem Detecting unusual events to prevent infrastructure damage.
method Categorizes six major OD methods: Statistical, Distance, Density, Clustering, Learning, and Ensemble.
result Advances and challenges of each method discussed.
To promote secure and private artificial intelligence (SPAI), we review studies on the model security and data privacy of DNNs. Model security allows system to behave as intended without being affected by malicious external influences that can compromise its integrity and efficiency. Security attacks can be divided bas…
New method improves black-box attacks using pre-trained models.
problem Efficiently attacking black-box models with limited information.
method EigenBA algorithm leveraging pre-trained white-box model's Jacobian matrix.
result Optimal perturbations are related to right singular vectors of Jacobian matrix.
RESTA defends LLMs against jailbreaking attacks by adding random noise to embeddings.
problem Vulnerability of LLMs to jailbreaking attacks that generate harmful outputs.
method Adds random noise to embedding vectors and aggregates during token generation.
result RESTA achieves superior robustness versus utility tradeoffs compared to baseline defenses.
As adversarial attacks pose a serious threat to the security of AI system in practice, such attacks have been extensively studied in the context of computer vision applications. However, few attentions have been paid to the adversarial research on automatic path finding. In this paper, we show dominant adversarial exam…
A new method to protect enterprise data privacy in AI models.
problem Enterprise data leakage risks in AI models.
method ABack, a training-free mechanism using Hidden State Model.
result Improves privacy utility by up to 15% over strong baselines.
Mathematical conditions and practical computations for adversarial robustness measures are established.
problem Existence, uniqueness, and scalability of adversarial robustness measures for AI classifiers.
method Formulated and proven mathematical conditions for existence, uniqueness, and explicit analytical computation of minimal adversarial paths and distances. Practical computation demonstrated on various AI tools and synthetic benchmarks.
result Explicit mathematical conditions and practical computations for adversarial robustness measures are established.
New framework evaluates MIA without retraining, addressing biases.
problem Evaluate MIA without retraining and under distribution shift.
method Causal inference approach to MIA evaluation.
result Practical estimators for MIA metrics without retraining.
The ever-growing big data and emerging artificial intelligence (AI) demand the use of machine learning (ML) and deep learning (DL) methods. Cybersecurity also benefits from ML and DL methods for various types of applications. These methods however are susceptible to security attacks. The adversaries can exploit the tra…
Survey on threats to federated learning models.
problem Vulnerabilities in federated learning protocols.
method Taxonomy of threat models and attacks.
result Important future research directions.
This research develops a new framework to measure AI investment returns considering both gains and risks.
problem Traditional ROI calculations fail to account for AI's dual impact on risk reduction and new exposures.
method Integrates ISO 42001 and regulatory exposure into a comprehensive financial framework using risk quantification methods.
result Accurate AI investment evaluation requires modeling both productivity gains and risk exposures.
During the last years, a remarkable breakthrough has been made in AI domain thanks to artificial deep neural networks that achieved a great success in many machine learning tasks in computer vision, natural language processing, speech recognition, malware detection and so on. However, they are highly vulnerable to easi…
Adversarial examples are considered a serious issue for safety critical applications of AI, such as finance, autonomous vehicle control and medicinal applications. Though significant work has resulted in increased robustness of systems to these attacks, systems are still vulnerable to well-crafted attacks. To address t…
Framework explains deep learning candlestick recognition.
problem Deep learning models explain candlestick patterns in a black box.
method Local search adversarial attacks to explain model reasoning.
result Model perceives candlestick patterns similarly to human traders.
Deep Learning based AI systems have shown great promise in various domains such as vision, audio, autonomous systems (vehicles, drones), etc. Recent research on neural networks has shown the susceptibility of deep networks to adversarial attacks - a technique of adding small perturbations to the inputs which can fool a…
Paper shows how to fool mammogram classifiers with adversarial attacks.
problem Vulnerability of mammographic image classifiers to adversarial attacks.
method Trained model on mamographic images, generated adversarial samples, analyzed similarity.
result Demonstrated successful adversarial attacks on mammographic image classifier.
The paper solves optimal bounds for separating data points in high dimensions.
problem Correcting AI errors and analyzing vulnerabilities in high-dimensional data.
method General stochastic separation theorems with optimal probability estimates.
result Explicit and optimal estimates of separation probabilities for important classes of distributions.
Paper tackles cybersecurity attack detection with an ensemble approach.
problem Challenges in multi-class classification for cyber security breaches.
method Designing a multi-node multi-class classification ensemble approach.
result Proposed approach outperforms full-data approach in multi-node data-censoring cases.
Characterizes deep neural network weight space for adversarial attacks.
problem Poor performance of deep learning models in adversarial examples.
method Characterizes deep neural network solution space using two paradigms.
result Adversarial attacks are less successful against Associative Memory Models.
Reprogram deep models to resist adversarial attacks without changing parameters.
problem Improving deep learning models' robustness against adversarial and noisy inputs.
method Proposes a non-linear robust pattern matching technique and three reprogramming paradigms.
result Demonstrates effective reprogramming of deep models for robustness without altering parameters.
With the increasing adoption of AI, inherent security and privacy vulnerabilities formachine learning systems are being discovered. One such vulnerability makes itpossible for an adversary to obtain private information about the types of instancesused to train the targeted machine learning model. This so-called model i…
Adversarial Robustness Toolbox (ART) is a Python library supporting developers and researchers in defending Machine Learning models (Deep Neural Networks, Gradient Boosted Decision Trees, Support Vector Machines, Random Forests, Logistic Regression, Gaussian Processes, Decision Trees, Scikit-learn Pipelines, etc.) agai…
Paper exposes vulnerabilities in interpreting machine learning models using adversarial attacks on PD plots.
problem Vulnerability of permutation-based interpretation methods, particularly PD plots, to adversarial attacks.
method Adversarial framework to manipulate black-box models and produce deceptive PD plots.
result It is possible to hide discriminatory behaviors in machine learning models through interpretation tools like PD plots.
We consider complexity of Deep Neural Networks (DNNs) and their associated massive over-parameterization. Such over-parametrization may entail susceptibility to adversarial attacks, loss of interpretability and adverse Size, Weight and Power - Cost (SWaP-C) considerations. We ask if there are methodical ways (regulariz…
There are two big unsolved mathematical questions in artificial intelligence (AI): (1) Why is deep learning so successful in classification problems and (2) why are neural nets based on deep learning at the same time universally unstable, where the instabilities make the networks vulnerable to adversarial attacks. We p…
Math proves deep learning unstable, despite stable neural networks existing.
problem Unstable neural networks in deep learning despite stable ones existing.
method Mathematical proof showing instability of current training procedures.
result Proven existence of stable and accurate neural networks with variable dimensions, but current algorithms cannot compute them.
Develops a method to interpret deep learning models by identifying key features.
problem Revealing the decision-making process of blackbox models from raw data to predictions.
method Adversarial attacks to localize discriminative features with statistical guarantees.
result Locally identified features are both biologically plausible and statistically significant.
AI stocks hedge against AI singularity's economic impact.
problem AI singularity's displacement of consumption.
method Developed an asset pricing model with incomplete markets.
result AI stocks command a premium due to market incompleteness.