All tracked items across vulnerabilities, news, research, incidents, and regulatory updates.
This paper addresses source-free domain adaptation (SFDA, a technique that adapts AI models to new datasets without accessing the original training data) for time-series data, such as sensor readings or activity logs. The authors argue that existing methods lack interpretability and may learn spurious patterns, so they propose PrEPoA, a framework that evaluates which parts of the time-series data the model considers important before fine-tuning it on the target domain. They demonstrate their approach works better than existing methods across five different real-world datasets.
LangChain, a framework for building applications powered by LLMs (large language models), had a serialization injection vulnerability (a flaw where specially crafted data can be misinterpreted as legitimate code during the conversion of objects to JSON format) in its toJSON() method. The vulnerability occurred because the method failed to properly escape objects containing 'lc' keys, which LangChain uses internally to mark serialized objects, allowing attackers to trick the system into treating malicious user data as legitimate LangChain objects when deserializing (converting back from JSON format).
LangChain, a framework for building AI agents and applications powered by large language models, had a serialization injection vulnerability (a flaw in how it converts data to stored formats) in its dumps() and dumpd() functions before versions 0.3.81 and 1.2.5. The functions failed to properly escape dictionaries containing 'lc' keys, which LangChain uses internally to mark serialized objects, allowing attackers to trick the system into treating user-supplied data as legitimate LangChain objects during deserialization (converting stored data back into usable form).
A vulnerability in Hugging Face Transformers GLM4 allows attackers to run harmful code on a system by tricking users into opening a malicious file or visiting a malicious webpage. The problem occurs because the software doesn't properly check data when loading model weights (the numerical values that make the AI work), allowing deserialization of untrusted data (converting unsafe external files into code the system will execute).
A vulnerability in Hugging Face Transformers' X-CLIP checkpoint conversion allows attackers to execute arbitrary code (running commands they choose on a system) by tricking users into opening malicious files or visiting malicious pages. The flaw occurs because the code doesn't properly validate checkpoint data before deserializing it (converting stored data back into usable objects), which lets attackers inject malicious code that runs with the same permissions as the application.
A vulnerability in Hugging Face Transformers' HuBERT convert_config function allows attackers to execute arbitrary code (RCE, or remote code execution, where an attacker runs commands on a system) by tricking users into converting a malicious checkpoint (a saved model file). The flaw occurs because the function doesn't properly validate user input before using it to run Python code.
Hugging Face Transformers (a popular library for working with AI language models) has a vulnerability in its SEW-D convert_config function that allows attackers to run arbitrary code (any commands they want) on a victim's computer. The flaw exists because the function doesn't properly check user input before using it to execute Python code, and an attacker can exploit this by tricking a user into converting a malicious checkpoint (a saved model file).
A vulnerability in Hugging Face Transformers (a popular AI library) allows attackers to run arbitrary code on a user's computer through a malicious checkpoint (a saved model file). The flaw exists in the convert_config function, which doesn't properly validate user input before executing it as Python code, meaning an attacker can trick a user into converting a malicious checkpoint to execute code with the user's permissions.
A vulnerability in Hugging Face Transformers (a popular library for working with AI language models) allows attackers to run arbitrary code on a computer by tricking users into opening malicious files or visiting malicious websites. The flaw occurs because the software doesn't properly check data when loading saved model checkpoints (files that store a model's learned parameters), which lets attackers execute code by sending untrusted data through deserialization (the process of converting stored data back into usable objects).
A vulnerability in Hugging Face Transformers' Transformer-XL model allows attackers to run arbitrary code (remote code execution) on a victim's computer by tricking them into opening a malicious file or visiting a malicious webpage. The flaw occurs because the software doesn't properly validate data when reading model files, allowing attackers to exploit the deserialization process (converting saved data back into objects that the program can use) to inject and execute malicious code.
A vulnerability in Hugging Face Transformers' Perceiver model allows attackers to run malicious code on a user's computer by tricking them into opening a malicious file or visiting a harmful webpage. The flaw happens because the software doesn't properly check data when loading model files, allowing untrusted code to be executed (deserialization of untrusted data, where a program reconstructs objects from stored data without verifying they're safe).
Tencent HunyuanDiT (an AI image generation model) has a remote code execution vulnerability in its model_resume function that allows attackers to run arbitrary code if a user opens a malicious file or visits a malicious page. The flaw stems from improper validation of user input during deserialization (converting data from storage format back into usable objects), allowing attackers to execute code with root-level privileges.
CVE-2025-63664 is a flaw in the GT Edge AI Platform (before version 2.0.10-dev) where incorrect access control in the /api/v1/conversations/*/messages API allows attackers without permission to view other users' message histories with AI agents. This is classified as improper access control (CWE-284, a category of security flaws where systems fail to properly restrict what users can access).
Fix: Update @langchain/core to version 0.3.80 or 1.1.8, and update langchain to version 0.3.37 or 1.2.3. According to the source: 'This issue has been patched in @langchain/core versions 0.3.80 and 1.1.8, and langchain versions 0.3.37 and 1.2.3.'
NVD/CVE DatabaseFix: Update to LangChain version 0.3.81 or version 1.2.5, where this issue has been patched.
NVD/CVE DatabaseThis research addresses machine unlearning in neural IR (information retrieval, the technology that ranks search results), a process called neural machine unranking (NuMuR) that selectively removes data from AI systems for privacy compliance. The authors propose CoCoL (contrastive and consistent loss, a method with two complementary training objectives), which uses a contrastive loss to reduce relevance scores on forgotten data while preserving performance on shared data, plus a consistent loss to maintain accuracy on retained data, demonstrating effective data removal across multiple neural ranking models.
Fix: The proposed solution is CoCoL, a dual-objective framework comprising: 1) a contrastive loss that reduces relevance scores on forget sets while maintaining performance on entangled samples, and 2) a consistent loss that preserves accuracy on the retain set. According to the paper, CoCoL achieves substantial forgetting with minimal retention and generalization performance loss.
IEEE Xplore (Security & AI Journals)Fix: Update GT Edge AI Platform to version 2.0.10-dev or later.
NVD/CVE DatabaseLarge language models (LLMs, AI systems trained on huge amounts of text to generate human-like responses) can now mimic not just general human language but also unusual, individual-specific human behaviors. This ability could lead to LLMs being used more widely in research studies and potentially reduce the role of actual humans, which raises concerns about AI alignment (ensuring AI systems behave in ways humans intend and approve of) and how this technology affects society.
Elderly people are increasingly using digital technology for communication and information access, but their limited cybersecurity knowledge makes them attractive targets for cybercriminals. The article examines common cybercrimes targeting seniors, the specific vulnerabilities that put them at risk, and existing approaches to reduce these dangers.
Generative AI (systems that create new text, images, or other content) is transforming many industries but raises ethical concerns like data privacy (protecting personal information), bias (unfair treatment of certain groups), transparency (being open about how the AI works), and accountability (responsibility for the AI's actions). Researchers propose a trust framework based on transparency, fairness, accountability, and privacy to help ensure generative AI is developed and used responsibly.
Cyberbullying on social media is a growing problem that harms people's mental health, and traditional methods to stop it are no longer effective. This study examines how artificial intelligence can help protect online communities from cyberbullying by exploring different AI technologies, their uses, and the challenges involved. The goal is to understand how AI might create safer online environments.
Pre-trained language models (PLMs, large AI systems trained on text data) can be vulnerable to backdoor attacks, where hidden triggers in input cause the model to produce manipulated output. This paper proposes an enhanced defense method called masking-differential prompting (MDP) that works with few-shot learning (training on very small datasets), using Jensen-Shannon divergence (a mathematical measure to compare probability distributions) instead of traditional methods and an automatic threshold-selection approach to better detect and block these attacks.
Fix: The paper proposes two enhancements to the masking-differential prompting (MDP) defense method: (1) adopting Jensen–Shannon (JS) divergence instead of Kullback–Leibler (KL) divergence to handle cases where anchor set information has insufficient density, keeping the divergence finite and better exploiting available data; and (2) proposing an adaptive threshold method that automatically searches for the threshold based on false rejection rate (FRR) allowance, replacing the computationally expensive manual threshold selection method using ROC curve (AUC).
IEEE Xplore (Security & AI Journals)This research addresses a problem in federated learning (a method where multiple computers train an AI model together without sharing raw data) combined with adversarial training (a technique that makes AI models resistant to intentionally tricky inputs). The authors found that simply combining these two approaches causes the model's accuracy to drop because adversarial training increases differences in the data across different computers, making the federated learning less effective. They propose SFAT (Slack Federated Adversarial Training), which uses a relaxation mechanism to adjust how the computers combine their learning results, reducing the harmful effects of data differences and improving overall performance.