Sequential Recommenders, which exploit dynamic user intents through interaction sequences, are vulnerable to adversarial attacks. While existing attacks primarily rely on data poisoning, they require large-scale user access or fake profiles thus lacking practicality. In this paper, we focus on the Profile Pollution Attack (PPA) that subtly contaminates partial user interactions to induce targeted mispredictions. Previous PPA methods suffer from two limitations, i.e., i) over-reliance on sequence horizon impact restricts fine-grained perturbations on item transitions, and ii) holistic modifications cause detectable distribution shifts. To address these challenges, we propose a constrained reinforcement driven attack CREAT that synergizes a bi-level optimization framework with multi-reward reinforcement learning to balance adversarial efficacy and stealthiness. We first develop a Pattern Balanced Rewarding Policy, which integrates pattern inversion rewards to invert critical patterns and distribution consistency rewards to minimize detectable shifts via unbalanced co-optimal transport. Then we employ a Constrained Group Relative Reinforcement Learning paradigm, enabling step-wise perturbations through dynamic barrier constraints and group-shared experience replay, achieving targeted pollution with minimal detectability. Extensive experiments demonstrate the effectiveness of CREAT.

Efficient and lightweight adaptation of pre-trained Vision-Language Models (VLMs) to downstream tasks through collaborative interactions between local clients and a central server is a rapidly emerging research topic in federated learning. Existing adaptation algorithms are typically trained iteratively, which incur significant communication costs and increase the susceptibility to potential attacks. Motivated by the one-shot federated training techniques that reduce client-server exchanges to a single round, developing a lightweight one-shot federated VLM adaptation method to alleviate these issues is particularly attractive. However, current one-shot approaches face certain challenges in adapting VLMs within federated settings: (1) insufficient exploitation of the rich multimodal information inherent in VLMs; (2) lack of specialized adaptation strategies to systematically handle the severe data heterogeneity; and (3) requiring additional training resource of clients or server. To bridge these gaps, we propose a novel Training-free One-shot Federated Adaptation framework for VLMs, named TOFA. To fully leverage the generalizable multimodal features in pre-trained VLMs, TOFA employs both visual and textual pipelines to extract task-relevant representations. In the visual pipeline, a hierarchical Bayesian model learns personalized, class-specific prototype distributions. For the textual pipeline, TOFA evaluates and globally aligns the generated local text prompts for robustness. An adaptive weight calibration mechanism is also introduced to combine predictions from both modalities, balancing personalization and robustness to handle data heterogeneity. Our method is training-free, not relying on additional training resources on either the client or server side. Extensive experiments across 9 datasets in various federated settings demonstrate the effectiveness of the proposed TOFA method.

Federated recommender systems enable collaborative model training while keeping user interaction data local and sharing only essential model parameters, thereby mitigating privacy risks. However, existing methods overlook a critical issue, i.e., the stable learning of a generalized item embedding throughout the federated recommender system training process. Item embedding plays a central role in facilitating knowledge sharing across clients. Yet, under the cross-device setting, local data distributions exhibit significant heterogeneity and sparsity, exacerbating the difficulty of learning generalized embeddings. These factors make the stable learning of generalized item embeddings both indispensable for effective federated recommendation and inherently difficult to achieve. To fill this gap, we propose a new federated recommendation framework, named Federated Recommendation with Generalized Embedding Learning (FedRecGEL). We reformulate the federated recommendation problem from an item-centered perspective and cast it as a multi-task learning problem, aiming to learn generalized embeddings throughout the training procedure. Based on theoretical analysis, we employ sharpness-aware minimization to address the generalization problem, thereby stabilizing the training process and enhancing recommendation performance. Extensive experiments on four datasets demonstrate the effectiveness of FedRecGEL in significantly improving federated recommendation performance. Our code is available at https://github.com/anonymifish/FedRecGEL.

Recommender systems have become increasingly influential in shaping user behavior and decision-making, highlighting their growing impact in various domains. Meanwhile, the widespread adoption of machine learning models in recommender systems has raised significant concerns regarding user privacy and security. As compliance with privacy regulations becomes more critical, there is a pressing need to address the issue of recommendation unlearning, i.e., eliminating the memory of specific training data from the learned recommendation models. Despite its importance, traditional machine unlearning methods are ill-suited for recommendation unlearning due to the unique challenges posed by collaborative interactions and model parameters. This survey offers a comprehensive review of the latest advancements in recommendation unlearning, exploring the design principles, challenges, and methodologies associated with this emerging field. We provide a unified taxonomy that categorizes different recommendation unlearning approaches, followed by a summary of widely used benchmarks and metrics for evaluation. By reviewing the current state of research, this survey aims to guide the development of more efficient, scalable, and robust recommendation unlearning techniques. Furthermore, we identify open research questions in this field, which could pave the way for future innovations not only in recommendation unlearning but also in a broader range of unlearning tasks across different machine learning applications.

With the growing demand for safeguarding sensitive user information in recommender systems, recommendation attribute unlearning is receiving increasing attention. Existing studies predominantly focus on single-attribute unlearning. However, privacy protection requirements in the real world often involve multiple sensitive attributes and are dynamic. Existing single-attribute unlearning methods cannot meet these real-world requirements due to CH1: the inability to handle multiple unlearning requests simultaneously, and CH2: the lack of efficient adaptability to dynamic unlearning needs. To address these challenges, we propose LEGO, a lightweight and efficient multiple-attribute unlearning framework. Specifically, we divide the multiple-attribute unlearning process into two steps: i) Embedding Calibration removes information related to a specific attribute from user embedding, and ii) Flexible Combination combines these embeddings into a single embedding, protecting all sensitive attributes. We frame the unlearning process as a mutual information minimization problem, providing LEGO a theoretical guarantee of simultaneous unlearning, thereby addressing CH1. With the two-step framework, where Embedding Calibration can be performed in parallel and Flexible Combination is flexible and efficient, we address CH2. Extensive experiments on three real-world datasets across three representative recommendation models demonstrate the effectiveness and efficiency of our proposed framework.

Although Multimodal Large Language Models (MLLMs) have advanced numerous fields, their training on extensive multimodal datasets introduces significant privacy concerns, prompting the necessity for efficient unlearning methods.However, current multimodal unlearning approaches often directly adapt techniques from unimodal contexts, largely overlooking the critical issue of modality alignment, i.e., consistently removing knowledge across both unimodal and multimodal settings. To close this gap, we introduce UMU-bench, a unified benchmark specifically targeting modality misalignment in multimodal unlearning. UMU-bench consists of a meticulously curated dataset featuring 653 individual profiles, each described with both unimodal and multimodal knowledge.Additionally, novel tasks and evaluation metrics focusing on modality alignment are introduced, facilitating a comprehensive analysis of unimodal and multimodal unlearning effectiveness. Through extensive experimentation with state-of-the-art unlearning algorithms on UMU-bench, we demonstrate prevalent modality misalignment issues in existing methods. These findings underscore the critical need for novel multimodal unlearning approaches explicitly considering modality alignment.

With the escalating privacy concerns in recommender systems, attribute unlearning has drawn widespread attention as an effective approach against attribute inference attacks. This approach focuses on unlearning users' privacy attributes to reduce the performance of attackers while preserving the overall effectiveness of recommendation. Current research attempts to achieve attribute unlearning through adversarial training and distribution alignment in the statistic setting. However, these methods often struggle in dynamic real-world environments, particularly when considering scenarios where unlearning requests are frequently updated. In this paper, we first identify three main challenges of current methods in dynamic environments, i.e., irreversible operation, low efficiency, and unsatisfied recommendation preservation. To overcome these challenges, we propose a Pluggable Attribute Unlearning framework, PAU. Upon receiving an unlearning request, PAU plugs an additional erasure module into the original model to achieve unlearning. This module can perform a reverse operation if the request is later withdrawn. To enhance the efficiency of unlearning, we introduce rate distortion theory and reduce the attack performance by maximizing the encoded bits required for users' embedding within the same class of the unlearned attribute and minimizing those for different classes, which eliminates the need to calculate the centroid distribution for alignment. We further preserve recommendation performance by constraining the compactness of the user embedding space around a reasonable flood level. Extensive experiments conducted on four real-world datasets and three mainstream recommendation models demonstrate the effectiveness of our proposed framework.

Recommender systems typically collect and analyze user data, which raises the risk of privacy invasion. User-sensitive information can be leaked from the user portrait, e.g., user embedding, within recommender models. Therefore, the task of recommendation unlearning has been widely studied, aiming to eliminate the influence of target data on recommender models. This paper explores the extended concept of unlearning, which seeks to remove sensitive user information while retaining the essential information for recommendation purposes. Previous studies have primarily focused on extended unlearning in isolation, e.g., attribute unlearning. However, users often need to fulfill multiple unlearning objectives simultaneously. Therefore, we bridge this gap by introducing post-training multi-objective unlearning, which allows the concurrent fulfillment of multiple unlearning objectives while preserving recommendation performance. Note that the objectives may conflict with each other, leading to the compromise of one objective when minimizing the overall objective value. To address this challenge, we introduce a Pareto exploration approach that incorporates the recommendation performance as optimization guidance, allowing us to obtain the Pareto optimal solution through the trade-off between conflicting objectives. To adapt to practical scenarios where data is not accessible post-training, we utilize a data-free regularization to guide recommendation performance. We conducted extensive experiments on three real-world datasets, which demonstrate the effectiveness of our proposed method.

Pre-trained language models acquire knowledge from vast amounts of text data, which can inadvertently contain sensitive information.To mitigate the presence of undesirable knowledge, the task of knowledge unlearning becomes crucial for language models.Previous research relies on gradient ascent methods to achieve knowledge unlearning, which is simple and effective.However, this approach calculates all the gradients of tokens in the sequence, potentially compromising the general ability of language models.To overcome this limitation, we propose an adaptive objective that calculates gradients with fine-grained control specifically targeting sensitive tokens.Our adaptive objective is pluggable, ensuring simplicity and enabling extension to the regularization-based framework that utilizes non-target data or other models to preserve general ability.Through extensive experiments targeting the removal of typical sensitive data, we demonstrate that our proposed method enhances the general ability of language models while achieving knowledge unlearning.Additionally, it demonstrates the capability to adapt to behavior alignment, eliminating all the undesirable knowledge within a specific domain.