arXiv subjects

Pengyuan Zhou

Publications and source records attributed to Pengyuan Zhou.

At least 37 records · Page 2Linked to original sources

arXiv2023

FedMKGC: Privacy-Preserving Federated Multilingual Knowledge Graph Completion

Knowledge graph completion (KGC) aims to predict missing facts in knowledge graphs (KGs), which is crucial as modern KGs remain largely incomplete. While training KGC models on multiple aligned KGs can improve performance, previous methods that rely on transferring raw data among KGs raise privacy concerns. To address this challenge, we propose a new federated learning framework that implicitly aggregates knowledge from multiple KGs without demanding raw data exchange and entity alignment. We treat each KG as a client that trains a local language model through textbased knowledge representation learning. A central server then aggregates the model weights from clients. As natural language provides a universal representation, the same knowledge thus has similar semantic representations across KGs. As such, the aggregated language model can leverage complementary knowledge from multilingual KGs without demanding raw user data sharing. Extensive experiments on a benchmark dataset demonstrate that our method substantially improves KGC on multilingual KGs, achieving comparable performance to state-of-the-art alignment-based models without requiring any labeled alignments or raw user data sharing. Our codes will be publicly available.

cs.CL↗

arXiv2023

User Authentication and Identity Inconsistency Detection via Mouse-trajectory Similarity Measurement

Completely Automated Public Turing Test To Tell Computers and Humans Apart (CAPTCHA) is a type of challenge-response test widely used in authentication systems. A well-known challenge it faces is the CAPTCHA farm, where workers are hired to solve CAPTCHAs manually. In this work, we propose to tackle this challenge from a novel perspective, converting CAPTCHA farm detection to identity inconsistency detection, which essentially becomes an authentication process. Specifically, we develop a novel embedding model, which measures the similarity between mouse trajectories collected during the session and when registering/solving CAPTCHA, to authenticate and detect identity inconsistency. Moreover, unlike most existing works that employ a separate mouse movement classifier for each individual user, which brings in considerable costs when serving a large number of users, our model performs detection tasks using only one classifier for all users, significantly reducing the cost. Experiment results validate the superiority of our method over the state-of-the-art time series classification methods, achieving 94.3% and 97.7% of AUC in identity and authentication inconsistency detection, respectively.

cs.CR↗

arXiv2023

AHSecAgg and TSKG: Lightweight Secure Aggregation for Federated Learning Without Compromise

Leveraging federated learning (FL) to enable cross-domain privacy-sensitive data mining represents a vital breakthrough to accomplish privacy-preserving learning. However, attackers can infer the original user data by analyzing the uploaded intermediate parameters during the aggregation process. Therefore, secure aggregation has become a critical issue in the field of FL. Many secure aggregation protocols face the problem of high computation costs, which severely limits their applicability. To this end, we propose AHSecAgg, a lightweight secure aggregation protocol using additive homomorphic masks. AHSecAgg significantly reduces computation overhead without compromising the dropout handling capability or model accuracy. We prove the security of AHSecAgg in semi-honest and active adversary settings. In addition, in cross-silo scenarios where the group of participants is relatively fixed during each round, we propose TSKG, a lightweight Threshold Signature based masking key generation method. TSKG can generate different temporary secrets and shares for different aggregation rounds using the initial key and thus effectively eliminates the cost of secret sharing and key agreement. We prove TSKG does not sacrifice security. Extensive experiments show that AHSecAgg significantly outperforms state-of-the-art mask-based secure aggregation protocols in terms of computational efficiency, and TSKG effectively reduces the computation and communication costs for existing secure aggregation protocols.

cs.CR↗

arXiv2023

Exploring User Perspectives on ChatGPT: Applications, Perceptions, and Implications for AI-Integrated Education

To foster the development of pedagogically potent and ethically sound AI-integrated learning landscapes, it is pivotal to critically explore the perceptions and experiences of the users immersed in these contexts. In this study, we perform a thorough qualitative content analysis across four key social media platforms. Our goal is to understand the user experience (UX) and views of early adopters of ChatGPT across different educational sectors. The results of our research show that ChatGPT is most commonly used in the domains of higher education, K-12 education, and practical skills training. In social media dialogues, the topics most frequently associated with ChatGPT are productivity, efficiency, and ethics. Early adopters' attitudes towards ChatGPT are multifaceted. On one hand, some users view it as a transformative tool capable of amplifying student self-efficacy and learning motivation. On the other hand, there is a degree of apprehension among concerned users. They worry about a potential overdependence on the AI system, which they fear might encourage superficial learning habits and erode students' social and critical thinking skills. This dichotomy of opinions underscores the complexity of Human-AI Interaction in educational contexts. Our investigation adds depth to this ongoing discourse, providing crowd-sourced insights for educators and learners who are considering incorporating ChatGPT or similar generative AI tools into their pedagogical strategies.

cs.CY↗

arXiv2023

NLPBench: Evaluating Large Language Models on Solving NLP Problems

Recent developments in large language models (LLMs) have shown promise in enhancing the capabilities of natural language processing (NLP). Despite these successes, there remains a dearth of research dedicated to the NLP problem-solving abilities of LLMs. To fill the gap in this area, we present a unique benchmarking dataset, NLPBench, comprising 378 college-level NLP questions spanning various NLP topics sourced from Yale University's prior final exams. NLPBench includes questions with context, in which multiple sub-questions share the same public information, and diverse question types, including multiple choice, short answer, and math. Our evaluation, centered on LLMs such as GPT-3.5/4, PaLM-2, and LLAMA-2, incorporates advanced prompting strategies like the chain-of-thought (CoT) and tree-of-thought (ToT). Our study reveals that the effectiveness of the advanced prompting strategies can be inconsistent, occasionally damaging LLM performance, especially in smaller models like the LLAMA-2 (13b). Furthermore, our manual assessment illuminated specific shortcomings in LLMs' scientific problem-solving skills, with weaknesses in logical decomposition and reasoning notably affecting results.

cs.CL↗

arXiv2023

Distilling Efficient Vision Transformers from CNNs for Semantic Segmentation

In this paper, we tackle a new problem: how to transfer knowledge from the pre-trained cumbersome yet well-performed CNN-based model to learn a compact Vision Transformer (ViT)-based model while maintaining its learning capacity? Due to the completely different characteristics of ViT and CNN and the long-existing capacity gap between teacher and student models in Knowledge Distillation (KD), directly transferring the cross-model knowledge is non-trivial. To this end, we subtly leverage the visual and linguistic-compatible feature character of ViT (i.e., student), and its capacity gap with the CNN (i.e., teacher) and propose a novel CNN-to-ViT KD framework, dubbed C2VKD. Importantly, as the teacher's features are heterogeneous to those of the student, we first propose a novel visual-linguistic feature distillation (VLFD) module that explores efficient KD among the aligned visual and linguistic-compatible representations. Moreover, due to the large capacity gap between the teacher and student and the inevitable prediction errors of the teacher, we then propose a pixel-wise decoupled distillation (PDD) module to supervise the student under the combination of labels and teacher's predictions from the decoupled target and non-target classes. Experiments on three semantic segmentation benchmark datasets consistently show that the increment of mIoU of our method is over 200% of the SoTA KD methods

cs.CV↗

arXiv2023

5G MEC Computation Handoff for Mobile Augmented Reality

The combination of 5G and Multi-access Edge Computing (MEC) can significantly reduce application delay by lowering transmission delay and bringing computational capabilities closer to the end user. Therefore, 5G MEC could enable excellent user experience in applications like Mobile Augmented Reality (MAR), which are computation-intensive, and delay and jitter-sensitive. However, existing 5G handoff algorithms often do not consider the computational load of MEC servers, are too complex for real-time execution, or do not integrate easily with the standard protocol stack. Thus they can impair the performance of 5G MEC. To address this gap, we propose Comp-HO, a handoff algorithm that finds a local solution to the joint problem of optimizing signal strength and computational load. Additionally, Comp-HO can easily be integrated into current LTE and 5G base stations thanks to its simplicity and standard-friendly deployability. Specifically, we evaluate Comp-HO through a custom NS-3 simulator which we calibrate via MAR prototype measurements from a real-world 5G testbed. We simulate both Comp-HO and several classic handoff algorithms. The results show that, even without a global optimum, the proposed algorithm still significantly reduces the number of large delays, caused by congestion at MECs, at the expense of a small increase in transmission delay.

cs.NI↗

arXiv2023

How Secure is Your Website? A Comprehensive Investigation on CAPTCHA Providers and Solving Services

Completely Automated Public Turing Test To Tell Computers and Humans Apart (CAPTCHA) has been implemented on many websites to identify between harmful automated bots and legitimate users. However, the revenue generated by the bots has turned circumventing CAPTCHAs into a lucrative business. Although earlier studies provided information about text-based CAPTCHAs and the associated CAPTCHA-solving services, a lot has changed in the past decade regarding content, suppliers, and solvers of CAPTCHA. We have conducted a comprehensive investigation of the latest third-party CAPTCHA providers and CAPTCHA-solving services' attacks. We dug into the details of CAPTCHA-As-a-Service and the latest CAPTCHA-solving services and carried out adversarial experiments on CAPTCHAs and CAPTCHA solvers. The experiment results show a worrying fact: most latest CAPTCHAs are vulnerable to both human solvers and automated solvers. New CAPTCHAs based on hard AI problems and behavior analysis are needed to stop CAPTCHA solvers.

cs.CR↗

arXiv2023

Detect Depression from Social Networks with Sentiment Knowledge Sharing

Social network plays an important role in propagating people's viewpoints, emotions, thoughts, and fears. Notably, following lockdown periods during the COVID-19 pandemic, the issue of depression has garnered increasing attention, with a significant portion of individuals resorting to social networks as an outlet for expressing emotions. Using deep learning techniques to discern potential signs of depression from social network messages facilitates the early identification of mental health conditions. Current efforts in detecting depression through social networks typically rely solely on analyzing the textual content, overlooking other potential information. In this work, we conduct a thorough investigation that unveils a strong correlation between depression and negative emotional states. The integration of such associations as external knowledge can provide valuable insights for detecting depression. Accordingly, we propose a multi-task training framework, DeSK, which utilizes shared sentiment knowledge to enhance the efficacy of depression detection. Experiments conducted on both Chinese and English datasets demonstrate the cross-lingual effectiveness of DeSK.

cs.CL↗

arXiv2023

Deepfake in the Metaverse: An Outlook Survey

We envision deepfake technologies, which synthesize realistic fake images and videos, will play an important role in the future metaverse. While enhancing users' immersion and experience with synthesized virtual characters and scenes, deepfake can cause serious consequences if used for fraud, impersonation, and dissemination of fake information. In this paper, we introduce the principles, applications, and risks of deepfake technology, and propose some countermeasures to help users and developers in the metaverse deal with the challenges brought by deepfake technologies. Further, we provide an outlook on the future development of deepfake in the metaverse.

cs.HC↗

arXiv2023

What if we have Meta GPT? From Content Singularity to Human-Metaverse Interaction in AIGC Era

The global metaverse development is facing a "cooldown moment", while the academia and industry attention moves drastically from the Metaverse to AI Generated Content (AIGC) in 2023. Nonetheless, the current discussion rarely considers the connection between AIGCs and the Metaverse. We can imagine the Metaverse, i.e., immersive cyberspace, is the black void of space, and AIGCs can simultaneously offer content and facilitate diverse user needs. As such, this article argues that AIGCs can be a vital technological enabler for the Metaverse. The article first provides a retrospect of the major pitfall of the metaverse applications in 2022. Second, we discuss from a user-centric perspective how the metaverse development will accelerate with AIGCs. Next, the article conjectures future scenarios concatenating the Metaverse and AIGCs. Accordingly, we advocate for an AI-Generated Metaverse (AIGM) framework for energizing the creation of metaverse content in the AIGC era.

cs.HC↗

arXiv2023

Spatiotemporal and Semantic Zero-inflated Urban Anomaly Prediction

Urban anomaly predictions, such as traffic accident prediction and crime prediction, are of vital importance to smart city security and maintenance. Existing methods typically use deep learning to capture the intra-dependencies in spatial and temporal dimensions. However, numerous key challenges remain unsolved, for instance, sparse zero-inflated data due to urban anomalies occurring with low frequency (which can lead to poor performance on real-world datasets), and both intra- and inter-dependencies of abnormal patterns across spatial, temporal, and semantic dimensions. Moreover, a unified approach to predict multiple kinds of anomaly is left to explore. In this paper, we propose STS to jointly capture the intra- and inter-dependencies between the patterns and the influential factors in three dimensions. Further, we use a multi-task prediction module with a customized loss function to solve the zero-inflated issue. To verify the effectiveness of the model, we apply it to two urban anomaly prediction tasks, crime prediction and traffic accident risk prediction, respectively. Experiments on two application scenarios with four real-world datasets demonstrate the superiority of STS, which outperforms state-of-the-art methods in the mean absolute error and the root mean square error by 37.88% and 18.10% on zero-inflated datasets, and, 60.32% and 37.28% on non-zero datasets, respectively.

cs.LG↗

arXiv2023

Vetaverse: A Survey on the Intersection of Metaverse, Vehicles, and Transportation Systems

Since 2021, the term "Metaverse" has been the most popular one, garnering a lot of interest. Because of its contained environment and built-in computing and networking capabilities, a modern car makes an intriguing location to host its own little metaverse. Additionally, the travellers don't have much to do to pass the time while traveling, making them ideal customers for immersive services. Vetaverse (Vehicular-Metaverse), which we define as the future continuum between vehicular industries and Metaverse, is envisioned as a blended immersive realm that scales up to cities and countries, as digital twins of the intelligent Transportation Systems, referred to as "TS-Metaverse", as well as customized XR services inside each Individual Vehicle, referred to as "IV-Metaverse". The two subcategories serve fundamentally different purposes, namely long-term interconnection, maintenance, monitoring, and management on scale for large transportation systems (TS), and personalized, private, and immersive infotainment services (IV). By outlining the framework of Vetaverse and examining important enabler technologies, we reveal this impending trend. Additionally, we examine unresolved issues and potential routes for future study while highlighting some intriguing Vetaverse services.

cs.HC↗

arXiv2023

FedACK: Federated Adversarial Contrastive Knowledge Distillation for Cross-Lingual and Cross-Model Social Bot Detection

Social bot detection is of paramount importance to the resilience and security of online social platforms. The state-of-the-art detection models are siloed and have largely overlooked a variety of data characteristics from multiple cross-lingual platforms. Meanwhile, the heterogeneity of data distribution and model architecture makes it intricate to devise an efficient cross-platform and cross-model detection framework. In this paper, we propose FedACK, a new federated adversarial contrastive knowledge distillation framework for social bot detection. We devise a GAN-based federated knowledge distillation mechanism for efficiently transferring knowledge of data distribution among clients. In particular, a global generator is used to extract the knowledge of global data distribution and distill it into each client's local model. We leverage local discriminator to enable customized model design and use local generator for data enhancement with hard-to-decide samples. Local training is conducted as multi-stage adversarial and contrastive learning to enable consistent feature spaces among clients and to constrain the optimization direction of local models, reducing the divergences between local and global models. Experiments demonstrate that FedACK outperforms the state-of-the-art approaches in terms of accuracy, communication efficiency, and feature space consistency.

cs.LG↗

arXiv2023

FRAS: Federated Reinforcement Learning empowered Adaptive Point Cloud Video Streaming

Point cloud video transmission is challenging due to high encoding/decoding complexity, high video bitrate, and low latency requirement. Consequently, conventional adaptive streaming methodologies often find themselves unsatisfactory to meet the requirements in threefold: 1) current algorithms reuse existing quality of experience (QoE) definitions while overlooking the unique features of point cloud video thus failing to provide optimal user experience, 2) most deep learning approaches require long-span data collections to learn sufficiently varied network conditions and result in long training periods and capacity occupation, 3) cloud training approaches pose privacy risks caused by leakage of user reported service usage and networking conditions. To overcome the limitations, we present FRAS, the first federated reinforcement learning framework, to the best of our knowledge, for adaptive point cloud video streaming. We define a new QoE model which takes the unique features of point cloud video into account. Each client uses reinforcement learning (RL) to train video quality selection with the objective of optimizing the user's QoE under multiple constraints. Then, a federated learning framework is integrated with the RL algorithm to enhance training performance with privacy preservation. Extensive simulations using real point cloud videos and network traces reveal the superiority of the proposed scheme over baseline schemes. We also implement a prototype that demonstrates the performance of FRAS via real-world tests.

cs.MM↗

arXiv2022

Celeritas: Fast Optimizer for Large Dataflow Graphs

The rapidly enlarging neural network models are becoming increasingly challenging to run on a single device. Hence model parallelism over multiple devices is critical to guarantee the efficiency of training large models. Recent proposals fall short either in long processing time or poor performance. Therefore, we propose Celeritas, a fast framework for optimizing device placement for large models. Celeritas employs a simple but efficient model parallelization strategy in the Standard Evaluation, and generates placement policies through a series of scheduling algorithms. We conduct experiments to deploy and evaluate Celeritas on numerous large models. The results show that Celeritas not only reduces the placement policy generation time by 26.4\% but also improves the model running time by 34.2\% compared to most advanced methods.

cs.DC↗

arXiv2022

Federated Split GANs

Mobile devices and the immense amount and variety of data they generate are key enablers of machine learning (ML)-based applications. Traditional ML techniques have shifted toward new paradigms such as federated (FL) and split learning (SL) to improve the protection of user's data privacy. However, these paradigms often rely on server(s) located in the edge or cloud to train computationally-heavy parts of a ML model to avoid draining the limited resource on client devices, resulting in exposing device data to such third parties. This work proposes an alternative approach to train computationally-heavy ML models in user's devices themselves, where corresponding device data resides. Specifically, we focus on GANs (generative adversarial networks) and leverage their inherent privacy-preserving attribute. We train the discriminative part of a GAN with raw data on user's devices, whereas the generative model is trained remotely (e.g., server) for which there is no need to access sensor true data. Moreover, our approach ensures that the computational load of training the discriminative model is shared among user's devices-proportional to their computation capabilities-by means of SL. We implement our proposed collaborative training scheme of a computationally-heavy GAN model in real resource-constrained devices. The results show that our system preserves data privacy, keeps a short training time, and yields same accuracy of model training in unconstrained devices (e.g., cloud). Our code can be found on https://github.com/YukariSonz/FSL-GAN

cs.LG↗

arXiv2022

HideNseek: Federated Lottery Ticket via Server-side Pruning and Sign Supermask

Federated learning alleviates the privacy risk in distributed learning by transmitting only the local model updates to the central server. However, it faces challenges including statistical heterogeneity of clients' datasets and resource constraints of client devices, which severely impact the training performance and user experience. Prior works have tackled these challenges by combining personalization with model compression schemes including quantization and pruning. However, the pruning is data-dependent and thus must be done on the client side which requires considerable computation cost. Moreover, the pruning normally trains a binary supermask $\in \{0, 1\}$ which significantly limits the model capacity yet with no computation benefit. Consequently, the training requires high computation cost and a long time to converge while the model performance does not pay off. In this work, we propose HideNseek which employs one-shot data-agnostic pruning at initialization to get a subnetwork based on weights' synaptic saliency. Each client then optimizes a sign supermask $\in \{-1, +1\}$ multiplied by the unpruned weights to allow faster convergence with the same compression rates as state-of-the-art. Empirical results from three datasets demonstrate that compared to state-of-the-art, HideNseek improves inferences accuracies by up to 40.6\% while reducing the communication cost and training time by up to 39.7\% and 46.8\% respectively.

cs.LG↗