arXiv subjects

Kehuan Zhang

Publications and source records attributed to Kehuan Zhang.

At least 19 recordsLinked to original sources

arXiv2026

From Multi-Agent to Single-Agent: When Is Skill Distillation Beneficial?

Multi-agent systems (MAS) for structured data-science tasks externalize analytical control through workflows spanning stages, tools, shared state, verification, and repair. Distilling such workflows into a single-agent skill can reduce orchestration overhead, but it remains unclear which workflow components should cross the control boundary. We distinguish capability resources, which expand what an agent can do, from pipeline guidance, which constrains which solutions it explores. On the same causal-estimation instances, adding task-qualified source pipeline guidance to a capability-matched skill changes normalized utility by +19.6 points under method-selection accuracy but -10.3 points under numerical error. To explain this reversal, we introduce Behavior-Outcome Freedom (F), a pre-synthesis diagnostic of signed behavior-outcome rank mismatch, and formalize its candidate-conditional role through Signed Anchor-Rank Transfer. Motivated by this mechanism, we propose AdaSkill, which preserves validated capability resources, removes runtime orchestration, and conditionally inherits pipeline guidance using a calibrated rule over F. Across 16 capability-matched interventions, the native-scale Full-minus-Discard effect decreases across the continuous F scale (r = -0.80, p < 0.001), while a 15-treatment atomic sweep localizes the reversal to pipeline guidance. Across 11 datasets spanning four structured data-science task families, AdaSkill combines strong task performance with substantially lower deployment overhead.

cs.AI↗

arXiv2026

When Agent Automation Becomes Profitable: Quantifying and Insuring Autonomous AI Risk through Trace-Economic Underwriting

AI agents can now take irreversible actions in operational systems, but agent-caused losses are still not clearly assigned, priced, or transferred. Providers often disclaim consequential damages, users are left with uncompensated losses, and default human review limits the efficiency gains of automation. We ask when autonomous AI deployment can become economically acceptable despite failure risk. Our answer is to quantify risk at the customer-task-trace episode level and transfer it through insurance. Automation is acceptable when its expected benefit exceeds the premium, control cost, and remaining risk. This requires a defined role with bounded permissions and comparable traces. We introduce trace-economic underwriting, which maps tool-use traces to customer exposure and claimable loss, then uses this representation for pricing, control, and risk transfer. It uses deterministic economic labels rather than an LLM judge. In our trace-to-loss testbed, trace-economic pricing reduces pricing MAE from $17.7K to $569 and removes regressive cross-subsidy. A 300-trace expert audit accepts 295 labels unchanged. On 1,000 real SWE-smith traces, trace-conditioned controls reduce CVaR95 by 72%. Theorem~1 gives a finite-sample scope condition. We release code, labels, and audit sheets.

cs.AI↗

arXiv2026

One Surrogate to Fool Them All: Universal, Transferable, and Targeted Adversarial Attacks with CLIP

Deep Neural Networks (DNNs) have achieved widespread success yet remain prone to adversarial attacks. Typically, such attacks either involve frequent queries to the target model or rely on surrogate models closely mirroring the target model -- often trained with subsets of the target model's training data -- to achieve high attack success rates through transferability. However, in realistic scenarios where training data is inaccessible and excessive queries can raise alarms, crafting adversarial examples becomes more challenging. In this paper, we present UnivIntruder, a novel attack framework that relies solely on a single, publicly available CLIP model and publicly available datasets. By using textual concepts, UnivIntruder generates universal, transferable, and targeted adversarial perturbations that mislead DNNs into misclassifying inputs into adversary-specified classes defined by textual concepts. Our extensive experiments show that our approach achieves an Attack Success Rate (ASR) of up to 85% on ImageNet and over 99% on CIFAR-10, significantly outperforming existing transfer-based methods. Additionally, we reveal real-world vulnerabilities, showing that even without querying target models, UnivIntruder compromises image search engines like Google and Baidu with ASR rates up to 84%, and vision language models like GPT-4 and Claude-3.5 with ASR rates up to 80%. These findings underscore the practicality of our attack in scenarios where traditional avenues are blocked, highlighting the need to reevaluate security paradigms in AI applications.

cs.CR↗

arXiv2026

Contextual Agentic Memory is a Memo, Not True Memory

Current agentic memory systems (vector stores, retrieval-augmented generation, scratchpads, and context-window management) do not implement memory: they implement lookup. We argue that treating lookup as memory is a category error with provable consequences for agent capability, long-term learning, and security. Retrieval generalizes by similarity to stored cases; weight-based memory generalizes by applying abstract rules to inputs never seen before. Conflating the two produces agents that accumulate notes indefinitely without developing expertise, face a provable generalization ceiling on compositionally novel tasks that no increase in context size or retrieval quality can overcome, and are structurally vulnerable to persistent memory poisoning as injected content propagates across all future sessions. Drawing on Complementary Learning Systems theory from neuroscience, we show that biological intelligence solved this problem by pairing fast hippocampal exemplar storage with slow neocortical weight consolidation, and that current AI agents implement only the first half. We formalize these limitations, address four alternative views, and close with a co-existence proposal and a call to action for system builders, benchmark designers, and the memory community.

cs.AI↗

arXiv2026

Reviewer Scores Are Not Comparable Across Research Areas in ML Peer Review

Peer review at ML conferences increasingly relies on reviewer scores as the primary decision instrument. As submissions have scaled from thousands to tens of thousands per year, no systematic audit has examined whether this instrument functions uniformly across research areas, or whether acceptance outcomes are in practice shaped by forces that reviewer scores neither capture nor control. This position paper argues that acceptance outcomes are shaped by forces beyond reviewer scores, and that the underlying cause is a measurement design failure, not individual bias. When a fixed numerical scale aggregates quality judgments across communities with structurally non-uniform reviewer pools, absolute scores become incomparable across areas, and area chairs must substitute community priors for score-based decisions. Using ICLR 2021--2026 data covering 50,289 papers across 219 research topics, we show that at any given reviewer score, a paper's acceptance probability varies by up to 8x depending on its topic. We rule out scoring culture, expert reviewer standards, rational area chair reweighting, and quality dilution as alternative explanations. We call on program committees to adopt inherently calibrated review signals and publish topic-stratified, score-conditional acceptance rates as a first-class fairness metric.

cs.DL↗

arXiv2026

LLM Agents Are Latent Context Managers: Eliciting Self-Managed Context via State Proprioception

Long-horizon tool agents are bottlenecked by how their context grows toward the limits of the context window. Recent systems make context management agent- or system-controlled, but they either learn compression policies that discard evidence or manage context in a layer the agent never sees. We argue that both miss a more basic gap: frontier language models are proprioceptively blind to their own context. From the prompt alone they cannot reliably infer block size, recency, or the remaining budget, all of which are needed for keep-or-archive decisions. We introduce VISTA (Visible Internal State for Tool Agents), a training-free, model-agnostic layer that represents working memory as typed addressable blocks, surfaces a runtime dashboard of token usage, recency, archive status, and remaining budget, and archives blocks as recoverable full-fidelity payloads. On LOCA-Bench, BrowseComp-Plus, and GAIA, the same untrained interface transfers across 1M-, 100K-, and 10K-scale trajectories. On LOCA-Bench it lifts Gemini-3-Flash from 22.7 to 50.7%, reaches 58.0% on BrowseComp-Plus, and remains competitive on GAIA. Gains grow with context pressure and transfer across backbones, while ablations confirm that the dashboard matters beyond archive and recovery tools.

cs.CL↗

arXiv2026

CLIP-Guided Backdoor Defense through Entropy-Based Poisoned Dataset Separation

Deep Neural Networks (DNNs) are susceptible to backdoor attacks, where adversaries poison training data to implant backdoor into the victim model. Current backdoor defenses on poisoned data often suffer from high computational costs or low effectiveness against advanced attacks like clean-label and clean-image backdoors. To address them, we introduce CLIP-Guided backdoor Defense (CGD), an efficient and effective method that mitigates various backdoor attacks. CGD utilizes a publicly accessible CLIP model to identify inputs that are likely to be clean or poisoned. It then retrains the model with these inputs, using CLIP's logits as a guidance to effectively neutralize the backdoor. Experiments on 4 datasets and 11 attack types demonstrate that CGD reduces attack success rates (ASRs) to below 1% while maintaining clean accuracy (CA) with a maximum drop of only 0.3%, outperforming existing defenses. Additionally, we show that clean-data-based defenses can be adapted to poisoned data using CGD. Also, CGD exhibits strong robustness, maintaining low ASRs even when employing a weaker CLIP model or when CLIP itself is compromised by a backdoor. These findings underscore CGD's exceptional efficiency, effectiveness, and applicability for real-world backdoor defense scenarios. Code: https://github.com/binyxu/CGD.

cs.MM↗

arXiv2026

Breaking the Stealth-Potency Trade-off in Clean-Image Backdoors with Generative Trigger Optimization

Clean-image backdoor attacks, which use only label manipulation in training datasets to compromise deep neural networks, pose a significant threat to security-critical applications. A critical flaw in existing methods is that the poison rate required for a successful attack induces a proportional, and thus noticeable, drop in Clean Accuracy (CA), undermining their stealthiness. This paper presents a new paradigm for clean-image attacks that minimizes this accuracy degradation by optimizing the trigger itself. We introduce Generative Clean-Image Backdoors (GCB), a framework that uses a conditional InfoGAN to identify naturally occurring image features that can serve as potent and stealthy triggers. By ensuring these triggers are easily separable from benign task-related features, GCB enables a victim model to learn the backdoor from an extremely small set of poisoned examples, resulting in a CA drop of less than 1%. Our experiments demonstrate GCB's remarkable versatility, successfully adapting to six datasets, five architectures, and four tasks, including the first demonstration of clean-image backdoors in regression and segmentation. GCB also exhibits resilience against most of the existing backdoor defenses.

cs.CV↗

arXiv2026

CAPED: Context-Aware Privacy Exposure Defense for Mobile GUI Agents

Screenshot-based mobile GUI agents can operate ordinary smartphone apps through the same visual interface as a human user, but this capability also turns every screen observation into a privacy boundary. During normal task execution, screenshots may expose contacts, messages, photos, files, recommendations, health cues, and other sensitive context that is unrelated to the user's request. We call this problem incidental visual privacy exposure. It is difficult to address with existing defenses: text anonymization misses many visual and inferential cues, while generic privacy masking can remove the evidence and controls that a GUI agent needs to complete the task. This paper presents CAPED, a context-aware pre-upload exposure control layer for mobile GUI agents. CAPED is designed as a phone-side protection layer: before screenshots are released to a remote multimodal agent, it extracts task requirements, uses screen context as a privacy prior, parses visible UI elements, and selectively exposes only content needed for the current task while masking incidental private content. We evaluate CAPED on AndroidWorld for broad task utility and with a controlled 28-task seeded privacy evaluation used as a measurement instrument for trajectory-level incidental leakage. In this seeded evaluation, Full CAPED reduces success-conditioned weighted seeded leakage from 0.766 under raw screenshots to 0.268 while preserving high task utility. A broader AndroidWorld run shows a remaining prototype-level utility cost, but the results show that task-driven selective exposure can reduce incidental visual leakage before screenshots are released to a remote GUI agent.

cs.CR↗

arXiv2026

From Internal Diagnosis to External Auditing: A VLM-Driven Paradigm for Data-Free Online Backdoor Defense

Deep Neural Networks remain inherently vulnerable to backdoor attacks. Traditional test-time defenses largely operate under the paradigm of internal diagnosis methods like model repairing or input robustness, yet these approaches are often fragile under advanced attacks as they remain entangled with the victim model's corrupted parameters. We propose a paradigm shift from Internal Diagnosis to External Semantic Auditing, arguing that effective defense requires decoupling safety from the victim model via an independent, semantically grounded auditor. To this end, we present a framework harnessing Universal Vision-Language Models (VLMs) as evolving semantic gatekeepers. We introduce PRISM (Prototype Refinement & Inspection via Statistical Monitoring), which overcomes the domain gap of general VLMs through two key mechanisms: a Hybrid VLM Teacher that dynamically refines visual prototypes online, and an Adaptive Router powered by statistical margin monitoring to calibrate gating thresholds in real-time. Extensive evaluation across 17 datasets and 11 attack types demonstrates that PRISM achieves state-of-the-art performance, suppressing Attack Success Rate to <1% on CIFAR-10 while improving clean accuracy, establishing a new standard for model-agnostic, externalized security.

cs.LG↗

arXiv2026

Trapping Attacker in Dilemma: Examining Internal Correlations and External Influences of Trigger for Defending GNN Backdoors

GNNs have become a standard tool for learning on relational data, yet they remain highly vulnerable to backdoor attacks. Prior defenses often depend on inspecting specific subgraph patterns or node features, and thus can be circumvented by adaptive attackers. We propose PRAETORIAN, a new defense that targets intrinsic requirements of effective GNN backdoors rather than surface-level cues. Our key observation is that flipping a victim node's prediction requires substantial influence on the victim: attackers tend to either inject many trigger nodes or rely on a small set of highly influential ones. Building on this observation, PRAETORIAN (i) analyzes internal correlations within potential trigger subgraphs to detect abnormally large injected structures, and (ii) quantifies external node influence to identify triggers with disproportionate impact. Across our evaluations, PRAETORIAN reduces the average attack success rate (ASR) to 0.55% with only a 0.62% drop in clean accuracy (CA), whereas state-of-the-art defenses still yield an average ASR of >20% and a CA drop of >3% under the same conditions. Moreover, PRAETORIAN remains effective against a range of adaptive attacks, forcing adversaries to either inject many trigger nodes to achieve high ASR (>80%), which incurs a >10% CA drop, or preserve CA at the cost of limiting ASR to 18.1%. Overall, PRAETORIAN constrains attackers to an unfavorable trade-off between efficacy and detectability.

cs.LG↗

arXiv2026

Beyond Nodes vs. Edges: A Multi-View Fusion Framework for Provenance-Based Intrusion Detection

Provenance-based intrusion detection has emerged as a promising approach for analyzing complex attack behaviors through system-level provenance graphs. However, existing defense methods face an inherent granularity limitation. Node-centric detectors, which evaluate anomalies using entities' attributes and local structural patterns, may misclassify benign behavioral changes or configuration modifications as suspicious. In contrast, edge-centric detectors, which focus more on interactions, may lack sufficient contextual awareness of the involved entities, leading to missed detections when compromised entities perform seemingly ordinary operations. These analytical biases highlight a persistent gap between node-centric and edge-centric analyses. To mitigate this gap, we present PROVFUSION, a multi-view detection framework that integrates anomaly signals from three distinct views (i.e., attribute, structure, and causality). The framework fuses heterogeneous anomaly signals through lightweight fusion schemes and determines the final anomaly decisions through a voting-based integration process, providing a more consistent and context-aware assessment of system behavior. This design enables PROVFUSION to capture both entity level deviations and interaction-level anomalies within a consistent analytic pipeline. Experiments on nine widely used benchmark datasets demonstrate that PROVFUSION achieves higher detection accuracy and lower false-positive rates than single node- and edge-centric baselines, maintaining stable performance across scenarios. Overall, the results suggest that our multi-view anomaly fusion together with voting-based decision aggregation offers a practical and effective direction for advancing provenance-based intrusion detection.

cs.CR↗

arXiv2025

Hunting the Ghost: Towards Automatic Mining of IoT Hidden Services

In this paper, we proposes an automatic firmware analysis tool targeting at finding hidden services that may be potentially harmful to the IoT devices. Our approach uses static analysis and symbolic execution to search and filter services that are transparent to normal users but explicit to experienced attackers. A prototype is built and evaluated against a dataset of IoT firmware, and The evaluation shows our tool can find the suspicious hidden services effectively.

cs.CR↗

arXiv2023

An Empirical Study of Blockchain System Vulnerabilities: Modules, Types, and Patterns

Blockchain, as a distributed ledger technology, becomes increasingly popular, especially for enabling valuable cryptocurrencies and smart contracts. However, the blockchain software systems inevitably have many bugs. Although bugs in smart contracts have been extensively investigated, security bugs of the underlying blockchain systems are much less explored. In this paper, we conduct an empirical study on blockchain's system vulnerabilities from four representative blockchains, Bitcoin, Ethereum, Monero, and Stellar. Specifically, we first design a systematic filtering process to effectively identify 1,037 vulnerabilities and their 2,317 patches from 34,245 issues/PRs (pull requests) and 85,164 commits on GitHub. We thus build the first blockchain vulnerability dataset. We then perform unique analyses of this dataset at three levels, including (i) file-level vulnerable module categorization by identifying and correlating module paths across projects, (ii) text-level vulnerability type clustering by natural language processing and similarity-based sentence clustering, and (iii) code-level vulnerability pattern analysis by generating and clustering code change signatures that capture both syntactic and semantic information of patch code fragments. Our analyses reveal three key findings: (i) some blockchain modules are more susceptible than the others; notably, each of the modules related to consensus, wallet, and networking has over 200 issues; (ii) about 70% of blockchain vulnerabilities are of traditional types, but we also identify four new types specific to blockchains; and (iii) we obtain 21 blockchain-specific vulnerability patterns that capture unique blockchain attributes and statuses, and demonstrate that they can be used to detect similar vulnerabilities in other popular blockchains, such as Dogecoin, Bitcoin SV, and Zcash.

cs.CR↗

arXiv2022

Nowhere to Hide: A Lightweight Unsupervised Detector against Adversarial Examples

Although deep neural networks (DNNs) have shown impressive performance on many perceptual tasks, they are vulnerable to adversarial examples that are generated by adding slight but maliciously crafted perturbations to benign images. Adversarial detection is an important technique for identifying adversarial examples before they are entered into target DNNs. Previous studies to detect adversarial examples either targeted specific attacks or required expensive computation. How design a lightweight unsupervised detector is still a challenging problem. In this paper, we propose an AutoEncoder-based Adversarial Examples (AEAE) detector, that can guard DNN models by detecting adversarial examples with low computation in an unsupervised manner. The AEAE includes only a shallow autoencoder but plays two roles. First, a well-trained autoencoder has learned the manifold of benign examples. This autoencoder can produce a large reconstruction error for adversarial images with large perturbations, so we can detect significantly perturbed adversarial examples based on the reconstruction error. Second, the autoencoder can filter out the small noise and change the DNN's prediction on adversarial examples with small perturbations. It helps to detect slightly perturbed adversarial examples based on the prediction distance. To cover these two cases, we utilize the reconstruction error and prediction distance from benign images to construct a two-tuple feature set and train an adversarial detector using the isolation forest algorithm. We show empirically that the AEAE is unsupervised and inexpensive against the most state-of-the-art attacks. Through the detection in these two cases, there is nowhere to hide adversarial examples.

cs.LG↗

arXiv2021

Towards Evaluating and Training Verifiably Robust Neural Networks

Recent works have shown that interval bound propagation (IBP) can be used to train verifiably robust neural networks. Reseachers observe an intriguing phenomenon on these IBP trained networks: CROWN, a bounding method based on tight linear relaxation, often gives very loose bounds on these networks. We also observe that most neurons become dead during the IBP training process, which could hurt the representation capability of the network. In this paper, we study the relationship between IBP and CROWN, and prove that CROWN is always tighter than IBP when choosing appropriate bounding lines. We further propose a relaxed version of CROWN, linear bound propagation (LBP), that can be used to verify large networks to obtain lower verified errors than IBP. We also design a new activation function, parameterized ramp function (ParamRamp), which has more diversity of neuron status than ReLU. We conduct extensive experiments on MNIST, CIFAR-10 and Tiny-ImageNet with ParamRamp activation and achieve state-of-the-art verified robustness. Code and the appendix are available at https://github.com/ZhaoyangLyu/VerifiablyRobustNN.

cs.CV↗

arXiv2020

Demon in the Variant: Statistical Analysis of DNNs for Robust Backdoor Contamination Detection

A security threat to deep neural networks (DNN) is backdoor contamination, in which an adversary poisons the training data of a target model to inject a Trojan so that images carrying a specific trigger will always be classified into a specific label. Prior research on this problem assumes the dominance of the trigger in an image's representation, which causes any image with the trigger to be recognized as a member in the target class. Such a trigger also exhibits unique features in the representation space and can therefore be easily separated from legitimate images. Our research, however, shows that simple target contamination can cause the representation of an attack image to be less distinguishable from that of legitimate ones, thereby evading existing defenses against the backdoor infection. In our research, we show that such a contamination attack actually subtly changes the representation distribution for the target class, which can be captured by a statistic analysis. More specifically, we leverage an EM algorithm to decompose an image into its identity part (e.g., person, traffic sign) and variation part within a class (e.g., lighting, poses). Then we analyze the distribution in each class, identifying those more likely to be characterized by a mixture model resulted from adding attack samples to the legitimate image pool. Our research shows that this new technique effectively detects data contamination attacks, including the new one we propose, and is also robust against the evasion attempts made by a knowledgeable adversary.

cs.CR↗

arXiv2019

Your Smart Home Can't Keep a Secret: Towards Automated Fingerprinting of IoT Traffic with Neural Networks

The IoT (Internet of Things) technology has been widely adopted in recent years and has profoundly changed the people's daily lives. However, in the meantime, such a fast-growing technology has also introduced new privacy issues, which need to be better understood and measured. In this work, we look into how private information can be leaked from network traffic generated in the smart home network. Although researchers have proposed techniques to infer IoT device types or user behaviors under clean experiment setup, the effectiveness of such approaches become questionable in the complex but realistic network environment, where common techniques like Network Address and Port Translation (NAPT) and Virtual Private Network (VPN) are enabled. Traffic analysis using traditional methods (e.g., through classical machine-learning models) is much less effective under those settings, as the features picked manually are not distinctive any more. In this work, we propose a traffic analysis framework based on sequence-learning techniques like LSTM and leveraged the temporal relations between packets for the attack of device identification. We evaluated it under different environment settings (e.g., pure-IoT and noisy environment with multiple non-IoT devices). The results showed our framework was able to differentiate device types with a high accuracy. This result suggests IoT network communications pose prominent challenges to users' privacy, even when they are protected by encryption and morphed by the network gateway. As such, new privacy protection methods on IoT traffic need to be developed towards mitigating this new issue.

cs.CR↗