With the rapid rise of social media and Internet culture, memes have become a popular medium for expressing emotional tendencies. This has sparked growing interest in Meme Emotion Understanding (MEU), which aims to classify the emotional intent behind memes by leveraging their multimodal contents. While existing efforts have achieved promising results, two major challenges remain: (1) a lack of fine-grained multimodal fusion strategies, and (2) insufficient mining of memes' implicit meanings and background knowledge. To address these challenges, we propose MemoDetector, a novel framework for advancing MEU. First, we introduce a four-step textual enhancement module that utilizes the rich knowledge and reasoning capabilities of Multimodal Large Language Models (MLLMs) to progressively infer and extract implicit and contextual insights from memes. These enhanced texts significantly enrich the original meme contents and provide valuable guidance for downstream classification. Next, we design a dual-stage modal fusion strategy: the first stage performs shallow fusion on raw meme image and text, while the second stage deeply integrates the enhanced visual and textual features. This hierarchical fusion enables the model to better capture nuanced cross-modal emotional cues. Experiments on two datasets, MET-MEME and MOOD, demonstrate that our method consistently outperforms state-of-the-art baselines. Specifically, MemoDetector improves F1 scores by 4.3% on MET-MEME and 3.4% on MOOD. Further ablation studies and in-depth analyses validate the effectiveness and robustness of our approach, highlighting its strong potential for advancing MEU.

While essential for stateful modern workloads like Large Language Model agents and IoT services, long-lived connections impede cloud infrastructure agility by complicating maintenance and load balancing. Existing connection migration solutions either lack support for industrial-grade encrypted traffic or fail to prevent packet loss during handover in active production environments. To address this gap, we propose Zephyr, a zero-loss and transparent TLS connection migration framework for cross-node migration between servers with different addresses. Zephyr ensures transport-layer consistency by orchestrating an eBPF-based packet buffering mechanism to safely intercept in-flight data. At the application layer, rather than deeply modifying standard TLS libraries, Zephyr creatively reuses the native session resumption mechanism via a “fake client” strategy to reconstruct complex cryptographic states without client involvement. Implemented in widely-used industrial stacks (Nginx and OpenSSL), Zephyr achieves connection migration with approximately 4.1 ms downtime and strict zero packet loss. This approach enables seamless infrastructure optimization without disrupting cloud services.

The transition to Data Processing Unit (DPU)-centric architectures has become the de-facto standard in modern cloud networks, enabling infrastructure offload and improved host resource utilization. However, the fixed hardware limits of DPUs increasingly fail to keep pace with the rapid growth of host compute density and network-intensive workloads. As a result, when DPU resources are saturated, host compute capacity often remains underutilized due to insufficient network provisioning.

Text revision is necessary to harness the written-text following human-acceptable requirements. Multi-intent text revision, however, requires all potential textual defects to be addressed in the same computational model, which poses a new challenge to the traditional single-intent-based text revision modeling approach. Conventional approaches often rely on models tailored to specific edit intents, limiting their ability to address diverse or unseen edit intents. Inspired by the reasoning strengths of Large Language Models (LLMs), we introduce EditCoT, a novel framework for multi-intent text revision. EditCoT breaks down the revision process into sequential reasoning steps, each targeting a specific text defect. The structured approach can enhance LLMs’ editing capabilities by enabling precise, intent-specific revisions within a unified model. We evaluate the effect of EditCoT on multi-/single-intent text revision tasks. For multi-intent tasks, EditCoT achieves state-of-the-art results, with a SARI score of 65.80 and a BERTScore of 88.27. For single-intent tasks, EditCoT, paired with GPT-o1, presents a competitive performance compared with specifically fine-tuned models. Furthermore, when combined with GPT-o1 or DeepSeek, EditCoT demonstrates impressive transferability to new edit intents via custom edit-chains. Overall, this study offers an effective framework for modeling and resolving text editing tasks, contributing a multi-intent dataset and an augmented single-intent dataset to support the community in advancing text revision research.

Large language models (LLMs) have demonstrated significant utility in a wide range of applications; however, their deployment is plagued by security vulnerabilities, notably jailbreak attacks. These attacks manipulate LLMs to generate harmful or unethical content by crafting adversarial prompts. While much of the current research on jailbreak attacks has focused on single-turn interactions, it has largely overlooked the impact of historical dialogues on model behavior. Although recent studies have explored multi-turn jailbreak attacks, they generally assume that the attacker can only manipulate the user prompt. In contrast, we highlight that an attacker can also control the model’s previous outputs. To this end, we introduce DIA, a new paradigm that leverages fabricated dialogue history to enhance jailbreak effectiveness. DIA operates in a black-box setting, requiring only access to the chat API or knowledge of the LLM’s chat template. We propose two methods for constructing adversarial historical dialogues: one adapts gray-box prefilling attacks, and the other exploits deferred responses. Our experiments demonstrate that DIA achieves state-of-the-art attack success rates on recent LLMs, including Llama-3.1 and GPT-4o. Additionally, we show that DIA can bypass 6 different defense mechanisms, highlighting its robustness.

The rapid expansion of cloud applications has led to unprecedented increases in network traffic volume, diversity, and complexity. As Cloud Service Providers (CSPs) adopt decentralized, geographically distributed data centers, effective traffic management across these environments has become critical. Distributed Rate Limiting (DRL) has emerged as an essential tool to manage the complex traffic dynamics of decentralized networks, yet traditional centralized rate limiting methods fall short, facing limitations in scalability, adaptability to bursty traffic, and efficiency. This paper presents C3PDAR (Cloud Control with Constant Probabilities and Dynamic Adjustment Range), a novel DRL algorithm tailored for decentralized cloud infrastructures. C3PDAR introduces three key innovations: (1) CPS-BPS DualPoint Rate Limiting and Parent-Child Token Bucket mechanisms, which effectively mitigate burst traffic and short-lived connections while improving bandwidth fairness and inter-tenant isolation; (2) A vSwitch-CGW Cascade Rate Limiting architecture, which reduces CPU overhead in CGW clusters and accelerates convergence by 42%–78%; (3) Virtual Extensible Local Area Network (VXLAN) Padding scheme, which embeds rate-limiting information in existing traffic instead of transmitting new data packets, reducing the communication overhead of the C3PDAR algorithm by over 40%. By integrating these advancements, C3PDAR delivers a scalable, robust solution that outperforms traditional DRL approaches in performance, fault tolerance, and resource efficiency. C3PDAR uniquely empowers CSPs to manage complex, high-volume traffic dynamics in decentralized cloud environments, offering both theoretical insights and practical optimizations for next-generation network control.

Deep learning (DL) frameworks serve as the backbone for a wide range of artificial intelligence applications. However, bugs within DL frameworks can cascade into critical issues in higher-level applications, jeopardizing reliability and security. While numerous techniques have been proposed to detect bugs in DL frameworks, research exploring common API patterns across frameworks and the potential risks they entail remains limited. Notably, many DL frameworks expose similar APIs with overlapping input parameters and functionalities, rendering them vulnerable to shared bugs, where a flaw in one API may extend to analogous APIs in other frameworks. To address this challenge, we propose MirrorFuzz, an automated API fuzzing solution to discover shared bugs in DL frameworks. MirrorFuzz operates in three stages: First, MirrorFuzz collects historical bug data for each API within a DL framework to identify potentially buggy APIs. Second, it matches each buggy API in a specific framework with similar APIs within and across other DL frameworks. Third, it employs large language models (LLMs) to synthesize code for the API under test, leveraging the historical bug data of similar APIs to trigger analogous bugs across APIs. We implement MirrorFuzz and evaluate it on four popular DL frameworks (TensorFlow, PyTorch, OneFlow, and Jittor). Extensive evaluation demonstrates that MirrorFuzz improves code coverage by 39.92% and 98.20% compared to state-of-the-art methods on TensorFlow and PyTorch, respectively. Moreover, MirrorFuzz discovers 315 bugs, 262 of which are newly found, and 80 bugs are fixed, with 52 of these bugs assigned CNVD IDs.

Most data-intensive applications currently run on NVMe storage, and virtualization is essential in cloud computing. Existing NVMe virtualization technologies include software-based and hardware-assisted. Virtio suffers from severe performance degradation, and polling-based solutions consume too many valuable CPU resources. Hardware-assisted solutions provide high performance and no CPU usage but have the challenges of developing dedicated hardware.In this paper, we propose NVMePass, a novel software-hardware co-design NVMe passthrough virtualization architecture designed to achieve high performance and no CPU overhead while maintaining high scalability. The key ideas of NVMePass are NVMe I/O queues passthrough for VMs and a mechanism to ensure security. The NVMePass supports DMA and interrupts remapping for VMs without hypervisor involvement, eliminating virtualization overhead and providing near-native performance. Isolation is achieved by I/O queues and logical block address resources exclusively allocated to VMs. We propose NVMe Resource Domain (NRD) and implement it in the NVMe controller to intercept illegal I/O requests. Thus, isolation and security are fully achieved. Results from our experiments show that NVMePass can provide comparable performance to VFIO, with an IOPS of $\mathbf{1 0 0. 1 \% - 1 0 0. 5 \%}$ of VFIO. Furthermore, compared to SPDK-Vhost, NVMePass achieves $\mathbf{4 0. 0 \%}$ lower latency when running 150 VMs, and NVMePass has an improvement of $\mathbf{6 8. 0 \%}$ OPS performance in a real-world application when running 100 VMs.

Widespread reuse of open-source code in smart contract development boosts programming efficiency but significantly amplifies bug propagation across contracts, while dedicated methods for detecting similar smart contract functions remain very limited. Conventional abstract-syntax-tree (AST) based methods for smart contract similarity detection face challenges in handling intricate tree structures, which impedes detailed semantic comparison of code. Recent deep-learning based approaches tend to overlook code syntax and detection interpretability, resulting in suboptimal performance. To fill this research gap, we introduceSmartDetector, a novel approach for computing similarity between smart contract functions, explainable at the fine-grained statement level. Technically,SmartDetectordecomposes the AST of a smart contract function into a series of smaller statement trees, each reflecting a structural element of the source code. Then,SmartDetectoruses a classifier to compute the similarity score of two functions by comparing each pair of their statement trees. To address the infinite hyperparameter space of the classifier, we mathematically derive a cosine-wise diffusion process to efficiently search optimal hyperparameters. Extensive experiments conducted on three large real-world datasets demonstrate thatSmartDetectoroutperforms current state-of-the-art methods by an average improvement of 14.01% in F1-score, achieving an overall average F1-score of 95.88%.

Differentially Private Stochastic Gradient Descent (DP-SGD) is a widely adopted technique for privacy-preserving deep learning. A critical challenge in DP-SGD is selecting the optimal clipping threshold C, which involves balancing the trade-off between clipping bias and noise magnitude, incurring substantial privacy and computing overhead during hyperparameter tuning. In this paper, we propose Dynamic Clipping DP-SGD (DC-SGD), a framework that leverages differentially private histograms to estimate gradient norm distributions and dynamically adjust the clipping thresholdC. Our framework includes two novel mechanisms: DC-SGD-P and DC-SGD-E. DC-SGD-P adjusts the clipping threshold based on a percentile of gradient norms, while DC-SGD-E minimizes the expected squared error of gradients to optimizeC. These dynamic adjustments significantly reduce the burden of hyperparameter tuningC. The extensive experiments on various deep learning tasks, including image classification and natural language processing, show that our proposed dynamic algorithms achieve up to 9 times acceleration on hyperparameter tuning than DP-SGD. And DC-SGD-E can achieve an accuracy improvement of 10.62% on CIFAR10 than DP-SGD under the same privacy budget of hyperparameter tuning. We conduct rigorous theoretical privacy and convergence analyses, showing that our methods seamlessly integrate with the Adam optimizer. Our results highlight the robust performance and efficiency of DC-SGD, offering a practical solution for differentially private deep learning with reduced computational overhead and enhanced privacy guarantees.

As cloud networks continue to evolve, network traffic has experienced an exponential increase. The network architecture is progressively adopting a distributed structure to address this challenge. This architecture extensively utilizes technologies like gateway clusters and Equal-Cost Multi-Path (ECMP) routing, enabling traffic from individual tenants to be routed through multiple pathways. As a result, distributed rate limiting (DRL) has emerged as an essential aspect. Nonetheless, the shift from centralized to DRL has encountered obstacles, with the associated algorithms grappling with simplicity, precision, and applicability issues. Consequently, our research seeks to reconceptualize the issue of DRL from a theoretical standpoint to discover a more holistic and efficacious solution.

Apsara vSwitch (AVS) is a per-host deployed forwarding component for instance network connectivity in the Alibaba Cloud. To meet the growing performance demands, we accelerated AVS by adopting the most widely used "Sep-path" offloading architecture, which introduces a separate hardware data path to speed up popular traffic. However, the deployment results prove that it is difficult to bridge the gap in performance and programming flexibility of the software and hardware data paths, resulting in unpredictable performance and low iteration velocity.

NVMe over Fabrics (NVMe-oF) has been widely applied as a remote storage protocol in cloud computing. The existing NVMe-oF software stack consumes a large number of CPU resources. Emerging devices, such as Smart NICs and DPUs, have supported hardware offloading of NVMe-oF to free these valuable CPU cores. However, NVMe-oF offloading capacity is always compromised because of limited hardware resources on design. Additionally, from thorough evaluations, we found that NVMe-oF inevitably suffers from severe performance degradation on complex application I/O patterns when using hardware offloading. It is challenging to achieve high performance and fully utilize NVMe-oF offloading simultaneously. In this paper, we propose HyQ, a novel hybrid I/O queue architecture for NVMe-oF, to achieve high performance while gaining the advantages of hardware offloading. HyQ realizes the coexistence of hardware offloading and software non-offloading queues, thus enabling the dynamic dispatching of I/O requests to appropriate processing queues according to user-defined I/O scheduling policies. Additionally, HyQ provides a request scheduling framework to support customized schedulers that select appropriate queues for I/O requests. In our evaluation, HyQ achieves up to 1.91x IOPS and 8.36x bandwidth performance improvement over the original hardware offloading scheme.

Differential Privacy (DP) is a widely used technique for protecting individuals' privacy by limiting what can be inferred about them from aggregate data. Recently, there have been efforts to implement DP using Secure Multi-Party Computation (MPC) to achieve high utility without the need for a trusted third party. One of the key components of implementing DP in MPC is noise sampling. Our work presents the first MPC solution for sampling discrete Gaussian, a common type of noise used for constructing DP mechanisms, which plays nicely with malicious secure MPC protocols.

Differential privacy (DP), as a rigorous mathematical definition quantifying privacy leakage, has become a well-accepted standard for privacy protection. Combined with powerful machine learning (ML) techniques, differentially private machine learning (DPML) is increasingly important. As the most classic DPML algorithm, DP-SGD incurs a significant loss of utility, which hinders DPML's deployment in practice. Many studies have recently proposed improved algorithms based on DP-SGD to mitigate utility loss. However, these studies are isolated and cannot comprehensively measure the performance of improvements proposed in algorithms. More importantly, there is a lack of comprehensive research to compare improvements in these DPML algorithms across utility, defensive capabilities, and generalizability.

Bare-metal instances are crucial for high-value, mission-critical applications on the cloud. Tenants exclusively use these dedicated hardware resources. Local virtualized disks are essential for bare-metal instances to provide flexible and high-performance storage resources. Traditionally tenants can choose polling-based software virtualization techniques, but they consume too many valuable host CPU cores and suffer from performance degradation. Cloud vendors are hard to deploy existing hardware-assisted local storage solutions in bare-metal instances due to no access to the host OS to install customized drivers. Moreover, cloud vendors have difficulties managing and maintaining the local storage devices in bare-metal instances because hardware resources and host operating systems are completely utilized by tenants, then it will impact the availability of storage devices.This paper presents our design and experience with BM-Store, a novel high-performance hardware-assisted virtual local storage architecture for bare-metal clouds. BM-Store is transparent to the host that tenants are unaware of the underlying hardware architecture. Therefore, it can be deployed on a large scale in cloud vendors. BM-Store consists of two components: an FPGA-based BMS-Engine and an ARM-based BMS-Controller. The BMS-Engine accelerates the I/O path to enable high-performance virtual storage independent of disk devices without consuming any CPU resource on the host. The BMS-Controller is responsible for resource management and maintenance to achieve flexible and high available local storage. The results of the extensive experiments show that BM-Store can achieve near-native performance, which only introduces about 3 µs extra latency and average 4.0% throughput overhead to native disks. Compared to SPDK vhost, BM-Store achieves an average bandwidth improvement of 15.7% in microbenchmark and a maximum throughput enhancement of 13.4% in real-world applications.

The vSwitch, as a critical component for Virtual Machine (VM) network connectivity in cloud environments, has prompted increasing attention towards its forwarding performance. While software optimization schemes have limitations in meeting the expanding network capacity demands [11, 12, 15, 17, 18], hardware offloading architectures leveraging SoC, FPGA, and ASIC have been proposed to transfer the match-action workload [1, 3, 6, 7, 13, 16], addressing the growing need for network capacity.

Cloud computing has witnessed tremendous growth, prompting enterprises to migrate to the cloud for reliable and on-demand computing. Within a single Virtual Private Cloud (VPC), the number of instances (such as VMs, bare metals, and containers) has reached millions, posing challenges related to supporting millions of instances with network location decoupling from the underlying hardware, high elastic performance, and high reliability. However, academic studies have primarily focused on specific issues like high-speed data plane and virtualized routing infrastructure, while existing industrial network technologies fail to adequately address these challenges.

Web search queries reveal extensive sensitive information about users’ interests and preferences to the search engines and eavesdroppers. Obfuscation-based private web search solutions automatically generate dummy queries and send the obfuscated queries to the search engine to hide users’ search intentions. Despite many obfuscation methods and tools have been developed, there is no practical system for evaluating their utility performance and the vulnerability against modern privacy attacks. In this article, we propose and develop OB-WSPES, a uniform evaluation system for obfuscation-based web search privacy, which allows researchers to conduct fair analysis and evaluation of existing or newly developed web search privacy protection/attack techniques. Leveraging OB-WSPES, we model the obfuscation activities and systematically implement and evaluate five obfuscation schemes and 10 modern web search attacks on the public AOL dataset. Our results demonstrate that, counter-intuitively, adding more fake queries to a user’s real data does not necessarily yield better privacy. The query utility of obfuscated queries declines with the increasing amount of dummy queries, while the application utility does not. We discuss the experimental results and point out the four important factors that affect the web search privacy and utility. Further, we propose possible directions for future research.

A large amount of valuable information resides in a decentralized attributed social graph, where each user locally maintains a limited view of the graph. However, there exists a conflicting requirement between publishing an attributed social graph and protecting the privacy of sensitive information contained in each user's local data. In this paper, we aim to collect and generate attributed social graphs in a decentralized manner while providing local differential privacy (LDP) for the collected data. Existing LDP-based synthetic graph generation methods either fail to preserve important graph properties (such as modularity and clustering coefficient) due to excessive noise injection or are unable to process attribute data, thus limiting their adoption and applicability. To overcome these weaknesses, we propose AsgLDP, a novel technique to generate privacy-preserving attributed graph data while satisfying LDP. AsgLDP preserves various graph properties through carefully designing the injected noise and estimating the joint distribution of attribute data. There are two key steps in AsgLDP: 1) collecting and generating graph data while satisfying LDP, and 2) optimizing the privacy-utility tradeoff of the generated data while preserving general graph properties such as the degree distribution, community structure and attribute distribution. Through theoretical analysis as well as experiments over 6 real-world datasets, we demonstrate the effectiveness of AsgLDP in preserving general graph properties such as degree distribution, community structure and attributed community search, while rigorously satisfying LDP. We also show that AsgLDP achieves a superior balance between utility and privacy as compared to the state-of-the-art approaches.