Research

Researchers discovered a new attack called FUBA (federated unlearning backdoor attack) that exploits a privacy feature in federated learning (a technique where multiple parties train an AI model together without sharing their raw data). The attack uses malicious unlearning requests, which are supposed to let participants remove their data from a trained model, to secretly inject backdoors (hidden harmful behaviors) into the model instead. The attack is difficult to detect because it hides from existing security defenses.

This research proposes FedFFTNet, a system for identifying which camera model took a photo by using federated learning (a technique where AI models train on data kept private across multiple devices rather than sharing raw data centrally). The system uses a lightweight deep learning architecture and a Laplacian-based patch selection strategy (focusing on sharp, detailed areas of images) to identify cameras while maintaining privacy, achieving very high accuracy rates on standard benchmark datasets.

This article presents MaxDiv, a technique for machine unlearning, which is the process of removing specific knowledge from an AI model after training to protect privacy, even when the original training data is no longer available. MaxDiv works by creating special synthetic data samples that have opposite characteristics to the data being forgotten, and it uses knowledge distillation (a technique where a model learns to replicate another model's behavior) to ensure important information isn't accidentally lost during the unlearning process.

This article reviews prompt engineering (the practice of designing inputs like questions or instructions to guide AI systems toward better responses) and analyzes its strengths, weaknesses, opportunities, and threats using a SWOT framework. The review covers how prompt engineering can improve interactions with large language models (advanced AI systems trained on vast amounts of text) across industries like healthcare and education, while also identifying challenges around maintaining accuracy and efficiency.

This paper presents RINNs (reparameterizable integral neural networks), a new type of AI model designed to run efficiently on mobile devices with limited computing power. The key innovation is a reparameterization strategy that converts the complex mathematical structure used during training into a simpler feed-forward structure (a straightforward sequence of processing steps) at inference time, allowing these models to achieve high accuracy (79.1%) while running very fast (0.87 milliseconds) on mobile hardware.

This paper presents Co-AttenDWG, a new method for detecting offensive content by combining text and images together. The approach uses coattention (a technique where two types of data pay attention to each other simultaneously), dimension-wise gating (a mechanism that selectively emphasizes important features at a detailed level), and expert fusion (combining predictions from multiple specialized models) to better understand how text and visual information relate to each other.

This research presents a federated learning (FL, a technique where multiple parties train an AI model together without sharing raw data) approach for learning the structure of dynamic Bayesian networks (DBN, a statistical model that represents relationships between variables over time) from distributed time series data. The method addresses challenges like data privacy and heterogeneity (when different parties' data follows different patterns), and provides mathematical proof that the approach reliably converges to good results, which hadn't been formally guaranteed before in this setting.

N/A -- The provided content is a navigation menu and feature listing from GitHub's website, not a security issue, vulnerability report, or technical problem related to AI/LLMs.

ATLAS Data v5.0.0 introduces a new "Technique Maturity" field that categorizes AI attack techniques based on evidence level, ranging from feasible (proven in research) to realized (used in actual attacks). The release adds 11 new techniques covering AI agent attacks like context poisoning (injecting false information into an AI system's memory), credential theft from AI configurations, and prompt injection (tricking an AI by hiding malicious instructions in its input), plus updates to existing techniques and case studies.

This article describes a maritime operational technology shipboard testbed, which is a controlled testing environment on a ship that mimics real maritime systems. The testbed allows cybersecurity researchers and professionals to safely study cyberattacks and test defensive strategies without risking actual ships or critical systems.

This research addresses challenges in asynchronous federated learning (AFL, a distributed machine learning approach where multiple devices train a model on their own data without sending raw data to a central server), specifically when devices have different types of objective functions and uneven data. The authors propose two main improvements: a staleness-aware aggregation mechanism (a method that reduces the influence of outdated updates from slower devices) and a dynamic learning rate schedule (an adaptive parameter that adjusts training speed based on how delayed each device's updates are) to improve model accuracy and stability in real-world environments where devices have different computing power and network speeds.

This research presents LipVor, an algorithm that mathematically verifies whether a trained neural network (a computer model with interconnected nodes that learns patterns) follows partial monotonicity constraints, which means outputs change predictably with certain inputs. The method works by testing the network at specific points and using mathematical properties to guarantee the network behaves correctly across its entire domain, potentially allowing neural networks to be used in critical applications like credit scoring where trustworthiness and predictable behavior are required.

Researchers discovered a type of backdoor attack (hidden malicious instructions planted in AI systems) on multiagent reinforcement learning systems, where one adversary agent uses its actions to trigger hidden failures in other agents' decision-making policies. Unlike previous attacks that assumed unrealistic direct control over what victims observe, this attack is more practical because it works through normal agent interactions in partially observable environments (where agents cannot always see what others are doing). The researchers developed a training method to help adversary agents efficiently trigger these backdoors with minimal suspicious actions.

This research addresses privacy risks in decentralized optimization (where multiple networked computers work together to solve a problem without a central coordinator) by proposing ZS-DDAPush, an algorithm that adds mathematical noise structures to protect sensitive node information during communication. The key innovation is that ZS-DDAPush achieves privacy protection while maintaining the accuracy and efficiency of the optimization process, avoiding the typical trade-offs seen in other privacy methods like differential privacy (adding statistical noise to protect individual data) or encryption (scrambling data so only authorized parties can read it).

Researchers developed TabExtractor, a tool that can steal tabular models (AI systems trained on spreadsheet-like data) without needing access to the original training data or knowing how the model was built. The attack works by creating synthetic data samples and using a special neural network architecture called a contrastive tabular transformer (CTT, a type of AI that learns by comparing similar and different examples) to reverse-engineer a clone of the victim model that performs almost as well as the original. This research shows that tabular models face serious security risks from extraction attacks.

This article describes BMMA-GPT, a biometric authentication system that uses multiple forms of identification (like fingerprints and facial recognition) together with mathematical optimization to improve security and speed. The system uses a dual-threshold approach (two decision points to verify identity) and can be tailored to different organizational needs, achieving high accuracy while keeping verification time under 1.5 seconds.

Machine unlearning allows AI models to forget the effects of specific training samples, but verifying whether this actually happened is difficult because existing checks (like backdoor attacks or membership inference attacks, which test if a model remembers data by trying to extract or manipulate it) can be fooled by a dishonest model provider who simply retrains the model to pass the test rather than truly unlearning. This paper proposes IndirectVerify, a formal verification method that uses pairs of connected samples (trigger samples that are unlearned and reaction samples that should be affected by that unlearning) with intentional perturbations (small changes to training data) to create indirect evidence that unlearning actually occurred, making it harder to fake.

AI systems used for important decisions often rely on empirical risk minimization (ERM, a training method that reduces prediction errors on known data) to build models, but these systems can suffer from unintentional bias, lack of transparency, and other risks. The EU has established Ethics Guidelines requiring trustworthy AI to meet seven key requirements, yet current ERM-based design prioritizes accuracy over trustworthiness. This article argues that developers need to balance four core objectives when designing AI systems: fairness (not discriminating against groups), privacy (protecting user data), robustness (resisting intentional attacks like fake news), and explainability (being transparent about how decisions are made).

This research proposes a new method for deploying cyber deception (defensive tricks to confuse attackers) in networks by combining deep reinforcement learning (a type of AI that learns by trial and error) with game theory that accounts for time delays. The method uses an algorithm called proximal policy optimization (PPO, a technique for training AI to make optimal decisions) to figure out where and when to place deception resources, and tests show it outperforms existing approaches in handling complex network attacks.