Zhuotao Liu

Hanbo Huang, Yiran Zhang, Hao Zheng et al.

Large language model (LLM) watermarking has shown promise in detecting AI-generated content and mitigating misuse, with prior work claiming robustness against paraphrasing and text editing. In this paper, we argue that existing evaluations are not sufficiently adversarial, obscuring critical vulnerabilities and overstating the security. To address this, we introduce the adaptive robustness radius, a formal metric that quantifies the worst-case resilience of watermarks against adaptive adversaries. By lifting the paraphrase space into a KL-divergence ball, we approximate this radius and theoretically demonstrate that optimizing the attack context and model parameters can significantly reduce the approximate radius, making watermarks highly vulnerable to paraphrase attacks. Leveraging this insight, we propose RLCracker, a reinforcement learning (RL)-based adaptive attack that erases watermark signals with limited watermarked examples and limited access to the detector. Despite weak supervision, it empowers a 3B model to achieve 98.5% removal success with minimal semantic shift on 1,500-token Unigram-marked texts after training on only 100 short samples. This performance dramatically exceeds 6.75% by GPT-4o and generalizes across five model sizes over ten watermarking schemes. Our code is available at https://github.com/OTT0-OTO/RLCracker.

Lepeng Zhao, Zhenhua Zou, Shuo Li et al.

Mobile Graphical User Interface (GUI) agents have demonstrated strong capabilities in automating complex smartphone tasks by leveraging multimodal large language models (MLLMs) and system-level control interfaces. However, this paradigm introduces significant privacy risks, as agents typically capture and process entire screen contents, thereby exposing sensitive personal data such as phone numbers, addresses, messages, and financial information. Existing defenses either reduce UI exposure, obfuscate only task-irrelevant content, or rely on user authorization, but none can protect task-critical sensitive information while preserving seamless agent usability. We propose an anonymization-based privacy protection framework that enforces the principle of available-but-invisible access to sensitive data: sensitive information remains usable for task execution but is never directly visible to the cloud-based agent. Our system detects sensitive UI content using a PII-aware recognition model and replaces it with deterministic, type-preserving placeholders (e.g., PHONE_NUMBER#a1b2c) that retain semantic categories while removing identifying details. A layered architecture comprising a PII Detector, UI Transformer, Secure Interaction Proxy, and Privacy Gatekeeper ensures consistent anonymization across user instructions, XML hierarchies, and screenshots, mediates all agent actions over anonymized interfaces, and supports narrowly scoped local computations when reasoning over raw values is necessary. Extensive experiments on the AndroidLab and PrivScreen benchmarks show that our framework substantially reduces privacy leakage across multiple models while incurring only modest utility degradation, achieving the best observed privacy-utility trade-off among existing methods. Code available at: https://github.com/one-step-beh1nd/gui_privacy_protection

Xiyuan Zhao, Xinhao Deng, Qi Li et al.

Website Fingerprinting (WF) attacks can effectively identify the websites visited by Tor clients via analyzing encrypted traffic patterns. Existing attacks focus on identifying different websites, but their accuracy dramatically decreases when applied to identify fine-grained webpages, especially when distinguishing among different subpages of the same website. WebPage Fingerprinting (WPF) attacks face the challenges of highly similar traffic patterns and a much larger scale of webpages. Furthermore, clients often visit multiple webpages concurrently, increasing the difficulty of extracting the traffic patterns of each webpage from the obfuscated traffic. In this paper, we propose Oscar, a WPF attack based on multi-label metric learning that identifies different webpages from obfuscated traffic by transforming the feature space. Oscar can extract the subtle differences among various webpages, even those with similar traffic patterns. In particular, Oscar combines proxy-based and sample-based metric learning losses to extract webpage features from obfuscated traffic and identify multiple webpages. We prototype Oscar and evaluate its performance using traffic collected from 1,000 monitored webpages and over 9,000 unmonitored webpages in the real world. Oscar demonstrates an 88.6% improvement in the multi-label metric Recall@5 compared to the state-of-the-art attacks.

Xinchi Han, Weihao Jiang, Yingming Mao et al.

Recent years have witnessed the growing deployment of optical circuit switches (OCS) in commercial GPU clusters (e.g., Google A3 GPU cluster) optimized for machine learning (ML) workloads. Such clusters adopt a three-tier leaf-spine-OCS topology, servers attach to leaf-layer electronic packet switches (EPSes); these leaf switches aggregate into spine-layer EPSes to form a Pod; and multiple Pods are interconnected via core-layer OCSes. Unlike EPSes, OCSes only support circuit-based paths between directly connected spine switches, potentially inducing a phenomenon termed routing polarization, which refers to the scenario where the bandwidth requirements between specific pairs of Pods are unevenly fulfilled through links among different spine switches. The resulting imbalance induces traffic contention and bottlenecks on specific leaf-to-spine links, ultimately reducing ML training throughput. To mitigate this issue, we introduce a leaf-centric paradigm to ensure traffic originating from the same leaf switch is evenly distributed across multiple spine switches with balanced loads. Through rigorous theoretical analysis, we establish a sufficient condition for avoiding routing polarization and propose a corresponding logical topology design algorithm with polynomial-time complexity. Large-scale simulations validate up to 19.27% throughput improvement and a 99.16% reduction in logical topology computation overhead compared to Mixed Integer Programming (MIP)-based methods.

Jianshuo Dong, Sheng Guo, Hao Wang et al.

Search agents connect LLMs to the Internet, enabling access to broader and more up-to-date information. However, unreliable search results may also pose safety threats to end users, establishing a new threat surface. In this work, we conduct two in-the-wild experiments to demonstrate both the prevalence of low-quality search results and their potential to misguide agent behaviors. To counter this threat, we introduce an automated red-teaming framework that is systematic, scalable, and cost-efficient, enabling lightweight and harmless safety assessments of search agents. Building on this framework, we construct the SafeSearch benchmark, which includes 300 test cases covering five categories of risks (e.g., misinformation and indirect prompt injection). Using this benchmark, we evaluate three representative search agent scaffolds, covering search workflow, tool-calling, and deep research, across 7 proprietary and 8 open-source backend LLMs. Our results reveal substantial vulnerabilities of LLM-based search agents: when exposed to unreliable websites, the highest ASR reached 90.5% for GPT-4.1-mini under a search workflow setting. Moreover, our analysis highlights the limited effectiveness of common defense practices, such as reminder prompting. This emphasizes the value of our framework in promoting transparency for safer agent development. Our codebase and test cases are publicly available: https://github.com/jianshuod/SafeSearch.

Zhenhua Zou, Sheng Guo, Qiuyang Zhan et al.

The evolution of Large Language Models (LLMs) has shifted mobile computing from App-centric interactions to system-level autonomous agents. Current implementations predominantly rely on a "Screen-as-Interface" paradigm, which inherits structural vulnerabilities and conflicts with the mobile ecosystem's economic foundations. In this paper, we conduct a systematic security analysis of state-of-the-art mobile agents using Doubao Mobile Assistant as a representative case. We decompose the threat landscape into four dimensions - Agent Identity, External Interface, Internal Reasoning, and Action Execution - revealing critical flaws such as fake App identity, visual spoofing, indirect prompt injection, and unauthorized privilege escalation stemming from a reliance on unstructured visual data. To address these challenges, we propose Aura, an Agent Universal Runtime Architecture for a clean-slate secure agent OS. Aura replaces brittle GUI scraping with a structured, agent-native interaction model. It adopts a Hub-and-Spoke topology where a privileged System Agent orchestrates intent, sandboxed App Agents execute domain-specific tasks, and the Agent Kernel mediates all communication. The Agent Kernel enforces four defense pillars: (i) cryptographic identity binding via a Global Agent Registry; (ii) semantic input sanitization through a multilayer Semantic Firewall; (iii) cognitive integrity via taint-aware memory and plan-trajectory alignment; and (iv) granular access control with non-deniable auditing. Evaluation on MobileSafetyBench shows that, compared to Doubao, Aura improves low-risk Task Success Rate from roughly 75% to 94.3%, reduces high-risk Attack Success Rate from roughly 40% to 4.4%, and achieves near-order-of-magnitude latency gains. These results demonstrate Aura as a viable, secure alternative to the "Screen-as-Interface" paradigm.

Jinzhu Yan, Haotian Xu, Zhuotao Liu et al.

The emerging programmable networks sparked significant research on Intelligent Network Data Plane (INDP), which achieves learning-based traffic analysis at line-speed. Prior art in INDP focus on deploying tree/forest models on the data plane. We observe a fundamental limitation in tree-based INDP approaches: although it is possible to represent even larger tree/forest tables on the data plane, the flow features that are computable on the data plane are fundamentally limited by hardware constraints. In this paper, we present BoS to push the boundaries of INDP by enabling Neural Network (NN) driven traffic analysis at line-speed. Many types of NNs (such as Recurrent Neural Network (RNN), and transformers) that are designed to work with sequential data have advantages over tree-based models, because they can take raw network data as input without complex feature computations on the fly. However, the challenge is significant: the recurrent computation scheme used in RNN inference is fundamentally different from the match-action paradigm used on the network data plane. BoS addresses this challenge by (i) designing a novel data plane friendly RNN architecture that can execute unlimited RNN time steps with limited data plane stages, effectively achieving line-speed RNN inference; and (ii) complementing the on-switch RNN model with an off-switch transformer-based traffic analysis module to further boost the overall performance. We implement a prototype of BoS using a P4 programmable switch as our data plane, and extensively evaluate it over multiple traffic analysis tasks. The results show that BoS outperforms state-of-the-art in both analysis accuracy and scalability.

Xuanqi Liu, Zhuotao Liu, Qi Li et al.

The escalating focus on data privacy poses significant challenges for collaborative neural network training, where data ownership and model training/deployment responsibilities reside with distinct entities. Our community has made substantial contributions to addressing this challenge, proposing various approaches such as federated learning (FL) and privacy-preserving machine learning based on cryptographic constructs like homomorphic encryption (HE) and secure multiparty computation (MPC). However, FL completely overlooks model privacy, and HE has limited extensibility (confined to only one data provider). While the state-of-the-art MPC frameworks provide reasonable throughput and simultaneously ensure model/data privacy, they rely on a critical non-colluding assumption on the computing servers, and relaxing this assumption is still an open problem. In this paper, we present Pencil, the first private training framework for collaborative learning that simultaneously offers data privacy, model privacy, and extensibility to multiple data providers, without relying on the non-colluding assumption. Our fundamental design principle is to construct the n-party collaborative training protocol based on an efficient two-party protocol, and meanwhile ensuring that switching to different data providers during model training introduces no extra cost. We introduce several novel cryptographic protocols to realize this design principle and conduct a rigorous security and privacy analysis. Our comprehensive evaluations of Pencil demonstrate that (i) models trained in plaintext and models trained privately using Pencil exhibit nearly identical test accuracies; (ii) The training overhead of Pencil is greatly reduced: Pencil achieves 10 ~ 260x higher throughput and 2 orders of magnitude less communication than prior art; (iii) Pencil is resilient against both existing and adaptive (white-box) attacks.

Qi Tan, Qi Li, Yi Zhao et al.

Federated Learning (FL) trains a black-box and high-dimensional model among different clients by exchanging parameters instead of direct data sharing, which mitigates the privacy leak incurred by machine learning. However, FL still suffers from membership inference attacks (MIA) or data reconstruction attacks (DRA). In particular, an attacker can extract the information from local datasets by constructing DRA, which cannot be effectively throttled by existing techniques, e.g., Differential Privacy (DP). In this paper, we aim to ensure a strong privacy guarantee for FL under DRA. We prove that reconstruction errors under DRA are constrained by the information acquired by an attacker, which means that constraining the transmitted information can effectively throttle DRA. To quantify the information leakage incurred by FL, we establish a channel model, which depends on the upper bound of joint mutual information between the local dataset and multiple transmitted parameters. Moreover, the channel model indicates that the transmitted information can be constrained through data space operation, which can improve training efficiency and the model accuracy under constrained information. According to the channel model, we propose algorithms to constrain the information transmitted in a single round of local training. With a limited number of training rounds, the algorithms ensure that the total amount of transmitted information is limited. Furthermore, our channel model can be applied to various privacy-enhancing techniques (such as DP) to enhance privacy guarantees against DRA. Extensive experiments with real-world datasets validate the effectiveness of our methods.

Hanbo Huang, Yihan Li, Bowen Jiang et al.

Privacy-sensitive users require deploying large language models (LLMs) within their own infrastructure (on-premises) to safeguard private data and enable customization. However, vulnerabilities in local environments can lead to unauthorized access and potential model theft. To address this, prior research on small models has explored securing only the output layer within hardware-secured devices to balance model confidentiality and customization. Yet this approach fails to protect LLMs effectively. In this paper, we discover that (1) query-based distillation attacks targeting the secured top layer can produce a functionally equivalent replica of the victim model; (2) securing the same number of layers, bottom layers before a transition layer provide stronger protection against distillation attacks than top layers, with comparable effects on customization performance; and (3) the number of secured layers creates a trade-off between protection and customization flexibility. Based on these insights, we propose SOLID, a novel deployment framework that secures a few bottom layers in a secure environment and introduces an efficient metric to optimize the trade-off by determining the ideal number of hidden layers. Extensive experiments on five models (1.3B to 70B parameters) demonstrate that SOLID outperforms baselines, achieving a better balance between protection and downstream customization.

Zhenhua Zou, Zhuotao Liu, Lepeng Zhao et al.

The rapid adoption of agentic AI, powered by large language models (LLMs), is transforming enterprise ecosystems with autonomous agents that execute complex workflows. Yet we observe several key security vulnerabilities in LLM-driven multi-agent systems (MASes): fragmented identity frameworks, insecure communication channels, and inadequate defenses against Byzantine agents or adversarial prompts. In this paper, we present the first systematic analysis of these emerging multi-agent risks and explain why the legacy security strategies cannot effectively address these risks. Afterwards, we propose BlockA2A, the first unified multi-agent trust framework that enables secure and verifiable and agent-to-agent interoperability. At a high level, BlockA2A adopts decentralized identifiers (DIDs) to enable fine-grained cross-domain agent authentication, blockchain-anchored ledgers to enable immutable auditability, and smart contracts to dynamically enforce context-aware access control policies. BlockA2A eliminates centralized trust bottlenecks, ensures message authenticity and execution integrity, and guarantees accountability across agent interactions. Furthermore, we propose a Defense Orchestration Engine (DOE) that actively neutralizes attacks through real-time mechanisms, including Byzantine agent flagging, reactive execution halting, and instant permission revocation. Empirical evaluations demonstrate BlockA2A's effectiveness in neutralizing prompt-based, communication-based, behavioral and systemic MAS attacks. We formalize its integration into existing MAS and showcase a practical implementation for Google's A2A protocol. Experiments confirm that BlockA2A and DOE operate with sub-second overhead, enabling scalable deployment in production LLM-based MAS environments.

Xinhao Deng, Xiyuan Zhao, Qilei Yin et al.

Website fingerprinting enables an eavesdropper to determine which websites a user is visiting over an encrypted connection. State-of-the-art website fingerprinting (WF) attacks have demonstrated effectiveness even against Tor-protected network traffic. However, existing WF attacks have critical limitations on accurately identifying websites in multi-tab browsing sessions, where the holistic pattern of individual websites is no longer preserved, and the number of tabs opened by a client is unknown a priori. In this paper, we propose ARES, a novel WF framework natively designed for multi-tab WF attacks. ARES formulates the multi-tab attack as a multi-label classification problem and solves it using the novel Transformer-based models. Specifically, ARES extracts local patterns based on multi-level traffic aggregation features and utilizes the improved self-attention mechanism to analyze the correlations between these local patterns, effectively identifying websites. We implement a prototype of ARES and extensively evaluate its effectiveness using our large-scale datasets collected over multiple months. The experimental results illustrate that ARES achieves optimal performance in several realistic scenarios. Further, ARES remains robust even against various WF defenses.

Mengyu Zhang, Zhuotao Liu, Jingwen Huang et al.

Privacy-preserving machine learning (PPML) is critical to ensure data privacy in AI. Over the past few years, the community has proposed a wide range of provably secure PPML schemes that rely on various cryptography primitives. However, when it comes to large language models (LLMs) with billions of parameters, the efficiency of PPML is everything but acceptable. For instance, the state-of-the-art solution for confidential LLM inference represents at least 10,000-fold slower performance compared to plaintext inference. The performance gap is even larger when the context length increases. In this position paper, we propose a novel framework named Agentic-PPML to make PPML in LLMs practical. Our key insight is to employ a general-purpose LLM for intent understanding and delegate cryptographically secure inference to specialized models trained on vertical domains. By modularly separating language intent parsing - which typically involves little or no sensitive information - from privacy-critical computation, Agentic-PPML completely eliminates the need for the LLMs to process the encrypted prompts, enabling practical deployment of privacy-preserving LLM-centric services.

Jingjing Xue, Sheng Sun, Min Liu et al.

To effectively manage and utilize massive distributed data at the network edge, Federated Learning (FL) has emerged as a promising edge computing paradigm across data silos. However, FL still faces two challenges: system heterogeneity (i.e., the diversity of hardware resources across edge devices) and statistical heterogeneity (i.e., non-IID data). Although sparsification can extract diverse submodels for diverse clients, most sparse FL works either simply assign submodels with artificially-given rigid rules or prune partial parameters using heuristic strategies, resulting in inflexible sparsification and poor performance. In this work, we propose Learnable Personalized Sparsification for heterogeneous Federated learning (FedLPS), which achieves the learnable customization of heterogeneous sparse models with importance-associated patterns and adaptive ratios to simultaneously tackle system and statistical heterogeneity. Specifically, FedLPS learns the importance of model units on local data representation and further derives an importance-based sparse pattern with minimal heuristics to accurately extract personalized data features in non-IID settings. Furthermore, Prompt Upper Confidence Bound Variance (P-UCBV) is designed to adaptively determine sparse ratios by learning the superimposed effect of diverse device capabilities and non-IID data, aiming at resource self-adaptation with promising accuracy. Extensive experiments show that FedLPS outperforms status quo approaches in accuracy and training costs, which improves accuracy by 1.28%-59.34% while reducing running time by more than 68.80%.

Xuanqi Liu, Zhuotao Liu

The community explored to build private inference frameworks for transformer-based large language models (LLMs) in a server-client setting, where the server holds the model parameters and the client inputs its private data (or prompt) for inference. However, these frameworks impose significant overhead when the private inputs are forward propagated through the original LLMs. In this paper, we show that substituting the computation- and communication-heavy operators in the transformer architecture with privacy-computing friendly approximations can greatly reduce the private inference costs while incurring very minor impact on model performance. Compared to state-of-the-art Iron (NeurIPS 2022), our privacy-computing friendly model inference pipeline achieves a $5\times$ acceleration in computation and an 80% reduction in communication overhead, while retaining nearly identical accuracy.

Jiaming Mu, Binghui Wang, Qi Li et al.

Graph Neural Networks (GNNs) have achieved state-of-the-art performance in various graph structure related tasks such as node classification and graph classification. However, GNNs are vulnerable to adversarial attacks. Existing works mainly focus on attacking GNNs for node classification; nevertheless, the attacks against GNNs for graph classification have not been well explored. In this work, we conduct a systematic study on adversarial attacks against GNNs for graph classification via perturbing the graph structure. In particular, we focus on the most challenging attack, i.e., hard label black-box attack, where an attacker has no knowledge about the target GNN model and can only obtain predicted labels through querying the target model.To achieve this goal, we formulate our attack as an optimization problem, whose objective is to minimize the number of edges to be perturbed in a graph while maintaining the high attack success rate. The original optimization problem is intractable to solve, and we relax the optimization problem to be a tractable one, which is solved with theoretical convergence guarantee. We also design a coarse-grained searching algorithm and a query-efficient gradient computation algorithm to decrease the number of queries to the target GNN model. Our experimental results on three real-world datasets demonstrate that our attack can effectively attack representative GNNs for graph classification with less queries and perturbations. We also evaluate the effectiveness of our attack under two defenses: one is well-designed adversarial graph detector and the other is that the target GNN model itself is equipped with a defense to prevent adversarial graph generation. Our experimental results show that such defenses are not effective enough, which highlights more advanced defenses.

Zhuotao Liu, Yangxi Xiang, Jian Shi et al.

Blockchain interoperability, which allows state transitions across different blockchain networks, is critical functionality to facilitate major blockchain adoption. Existing interoperability protocols mostly focus on atomic token exchange between blockchains. However, as blockchains have been upgraded from passive distributed ledgers into programmable state machines (thanks to smart contracts), the scope of blockchain interoperability goes beyond just token exchange. In this paper, we present HyperService, the first platform that delivers interoperability and programmability across heterogeneous blockchains. HyperService is powered by two innovative designs: (i) a developer-facing programming framework that allows developers to build cross-chain applications in a unified programming model; and (ii) a secure blockchain-facing cryptography protocol that provably realizes those applications on blockchains. We implement a prototype of HyperService in about 35,000 lines of code to demonstrate its practicality. Our experiment results show that HyperService imposes reasonable latency, in order of seconds, on the end-to-end execution of cross-chain applications

Zhuotao Liu, Yuan Cao, Xuewu Zhang et al.

With the advent of software-defined networking, network configuration through programmable interfaces becomes practical, leading to various on-demand opportunities for network routing update in multi-tenant datacenters, where tenants have diverse requirements on network routings such as short latency, low path inflation, large bandwidth, high reliability, etc. Conventional solutions that rely on topology search coupled with an objective function https:// www.overleaf.com/project/5beb742041ab9c0e3caec84f to find desired routings have at least two shortcomings: (i) they run into scalability issues when handling consistent and frequent routing updates and (ii) they restrict the flexibility and capability to satisfy various routing requirements. To address these issues, this paper proposes a novel search and optimization decoupled design, which not only saves considerable topology search costs via search result reuse, but also avoids possible sub-optimality in greedy routing search algorithms by making decisions based on the global view of all possible routings. We implement a prototype of our proposed system, OpReduce, and perform extensive evaluations to validate its design goals.

Zhuotao Liu, Yuan Cao, Min Zhu et al.

Defending against distributed denial of service (DDoS) attacks in the Internet is a fundamental problem. However, recent industrial interviews with over 100 security experts from more than ten industry segments indicate that DDoS problems have not been fully addressed. The reasons are twofold. On one hand, many academic proposals that are provably secure witness little real-world deployment. On the other hand, the operation model for existing DDoS-prevention service providers (e.g., Cloudflare, Akamai) is privacy invasive for large organizations (e.g., government). In this paper, we present Umbrella, a new DDoS defense mechanism enabling Internet Service Providers (ISPs) to offer readily deployable and privacy-preserving DDoS prevention services to their customers. At its core, Umbrella develops a multi-layered defense architecture to defend against a wide spectrum of DDoS attacks. In particular, the flood throttling layer stops amplification-based DDoS attacks; the congestion resolving layer, aiming to prevent sophisticated attacks that cannot be easily filtered, enforces congestion accountability to ensure that legitimate flows are guaranteed to receive their fair shares regardless of attackers' strategies; and finally the userspecific layer allows DDoS victims to enforce self-desired traffic control policies that best satisfy their business requirements. Based on Linux implementation, we demonstrate that Umbrella is capable to deal with large scale attacks involving millions of attack flows, meanwhile imposing negligible packet processing overhead. Further, our physical testbed experiments and large scale simulations prove that Umbrella is effective to mitigate various DDoS attacks.

Zhuotao Liu, Yushan Liu, Philipp Winter et al.

Tor is the most widely used anonymity network, currently serving millions of users each day. However, there is no access control in place for all these users, leaving the network vulnerable to botnet abuse and attacks. For example, criminals frequently use exit relays as stepping stones for attacks, causing service providers to serve CAPTCHAs to exit relay IP addresses or blacklisting them altogether, which leads to severe usability issues for legitimate Tor users. To address this problem, we propose TorPolice, the first privacy-preserving access control framework for Tor. TorPolice enables abuse-plagued service providers such as Yelp to enforce access rules to police and throttle malicious requests coming from Tor while still providing service to legitimate Tor users. Further, TorPolice equips Tor with global access control for relays, enhancing Tor's resilience to botnet abuse. We show that TorPolice preserves the privacy of Tor users, implement a prototype of TorPolice, and perform extensive evaluations to validate our design goals.