Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks59 from public data
- STRIP1,145
This work builds STRong Intentional Perturbation (STRIP) based run-time trojan attack detection system and focuses on vision system, which achieves an overall false acceptance rate (FAR) of less than 1%, given a preset false rejection rate (FRR) of 1%, for different types of triggers.
- Android HIV: A Study of Repackaging Malware for Evading Machine-Learning Detection299
A new attacking method is introduced that generates adversarial examples of Android malware and evades being detected by the current models, and can also deceive recent machine learning-based detectors that rely on semantic features such as control-flow-graph.
- 88
- 88
- 65
- 61
- Man-in-the-Middle Attacks Against Machine Learning Classifiers Via Malicious Generative Models46
This work explores vulnerabilities of DNN models under the umbrella of Man-in-the-Middle (MitM) attacks, which has not been investigated before, and investigates using a VAE decoder to either transform benign inputs to their adversarial counterparts or decode outputs from benign VAE encoders to be adversarial examples.
- Defending Against Adversarial Attack Towards Deep Neural Networks Via Collaborative Multi-Task Training40
A novel defensive framework based on collaborative multi-task training, aiming at providing defence for different types of attacks is proposed, which can prevent adversaries from finding valid adversarial examples when the defence strategy is exposed.
- StyleFool: Fooling Video Classification Systems via Style Transfer35
StyleFool is proposed, a black-box video adversarial attack via style transfer to fool the video classification system and outperforms the state-of-the-art adversarial attacks in terms of both the number of queries and the robustness against existing defenses.
- QUEEN: Query Unlearning Against Model Extraction31
QUEEN (QUEry unlEarNing) is proposed that proactively launches counterattacks on potential model extraction attacks from the very beginning and outperforms the state-of-the-art defenses against various model extraction attacks with a relatively low cost to the model accuracy.
- Provably Unlearnable Data Examples20
A mechanism for certifying the so-called $(q, \eta)$-Learnability of an unlearnable dataset via parametric smoothing is proposed and a simple recovery attack can restore the clean-task performance of the classifiers trained on UEs by slightly perturbing the learned weights.
- Mitigating Unauthorized Speech Synthesis for Voice Protection16
An effective, transferable, and robust proactive protection technology named Pivotal Objective Perturbation (POP) that applies imperceptible error-minimizing noises on original speech samples to prevent them from being effectively learned for text-to-speech (TTS) synthesis models so that high-quality deepfake speeches cannot be generated.
- SafeSpeech: Robust and Universal Voice Protection Against Malicious Speech Synthesis13
This work proposes a defensive framework, SafeSpeech, which protects the users' audio before uploading by embedding imperceptible perturbations on original speeches to prevent high-quality synthetic speech.
- 11
- 11
- AI Model Modulation with Logits Redistribution9
This work proposes AIM, a novel model modulation paradigm that enables a single model to exhibit diverse behaviors to meet the specific end requirements, and introduces a logits redistribution strategy that operates in a training data-agnostic and retraining-free manner.
- Whispering Under the Eaves: Protecting User Privacy Against Commercial and LLM-powered Automatic Speech Recognition Systems9
Comprehensive evaluation on four commercial ASR APIs demonstrates the protection superiority of AudioShield over existing competitors, and both objective and subjective evaluations indicate that AudioShield significantly improves the audio quality.
- Re-Key-Free, Risky-Free: Adaptable Model Usage Control8
Experiments demonstrate that AdaLoc can offer a practical solution for adaptive and protected DNN deployment in evolving real-world scenarios, and establishes a formal foundation to underpin AdaLoc, providing crucial bounds such as the errors introduced by updates restricted to the access key.
- 8
- 8
- Towards Evaluating the Robustness of Automatic Speech Recognition Systems via Audio Style Transfer8
Experimental results show that the proposed attack on ASR systems based on user-customized style transfer can meet the need for user-customized styles and achieve a success rate of 82% in attacks, while keeping sound naturalness due to the authors' user study.
- Explore the hidden treasure in protein–protein interaction networks — An iterative model for predicting protein functions8
The iterative approach initially proposes an idea that iteratively incorporates the interaction information of unannotated proteins into the protein function prediction and can be applied on existing prediction algorithms to improve prediction performance.
- CyberAlly: Leveraging LLMs and Knowledge Graphs to Empower Cyber Defenders7
CyberAlly is introduced, a knowledge graph-enhanced AI assistant designed to enhance the efficiency and effectiveness of Blue Teams during incident response and equipping defenders to tackle evolving threats with greater precision and confidence.
- 7
- 6
- 6
- Discover Diamonds-in-the-Rough Using Interactive Visual Analytics System-Tweets as a Collective Diary of the Occupy Movement5
This case study demonstrates the combination of computational methods (e.g., topic modeling and event detection) and interactive visual analytics in facilitating users to examine how relevant tweets can reflect a collective view of a social movement by analyzing events extracted from the content of the tweets.
- Catch Me If You Can: Detecting Compromised Users Through Partial Observation on Networks5
A microscopic mathematical model is proposed to present the risk propagation from the exposed sources and shows a new finding that the ones who synchronously receive the risk copies from all monitored users are most likely to be the sources.
- 4
- 4
- Rethinking the Threat and Accessibility of Adversarial Attacks against Face Recognition Systems4
This paper proposes an effective yet easy-to-launch physical adversarial attack, named AdvColor, against black-box face recognition pipelines in the physical world, and conducts a survey on the threats of prevailing adversarial attacks, to understand the gap between the machine-measured and human-assessed threat levels of different forms of adversarial attacks.
- On the Neural Backdoor of Federated Generative Models in Edge Computing4
This article studies the vulnerabilities of federated GMs to data-poisoning-based backdoor attacks via gradient uploading and enhances the attack to reduce the required poisonous data samples and cope with dynamic network environments.
- How Spam Features Change in Twitter and the Impact to Machine Learning Based Detection4
The results suggested a counter-intuitive conclusion that the ‘Spam Drift’ problem does not have serious impact on spam detection Precision and non-spam detection Recall, two metrics that industries prioritise in practice.
- Query-Efficient Video Adversarial Attack With Stylized Logo on Service Computing3
Sufficient experimental results indicate that SLA can achieve better performance than state-of-the-art methods and still maintain good deception effects when facing various defense methods and serve as a memorandum of possible attack methods for the future.
- CAMP in the Odyssey: Provably Robust Reinforcement Learning with Certified Radius Maximization3
A novel paradigm dubbed CAMP is introduced, designed to enhance DRL policies, achieving better utility without compromising provable robustness, and introduces \textit{policy imitation} as a novel technique to stabilize training.
- 3
- Who Spread to Whom? Inferring Online Social Networks with User Features3
This paper developed an innovative method to address the inference inaccuracy caused by the exposed two problems, and determined the existence of an edge between a pair of users according to part of the users' features.
- STC: Exposing Hidden Compromised Devices in Networked Sustainable Green Smart Computing Platforms by Partial Observation3
An unsupervised classifier based on source-tracing technique (STC in short) to expose hidden compromised devices with partial observation on the networked sustainable green smart computing platforms is proposed.
- Bones of Contention: Exploring Query-Efficient Attacks Against Skeleton Recognition Systems2
Two query-effIcient Skeletal Adversarial AttaCks, ISAAC-K and ISAAC-N are proposed to improve the robustness of skeleton recognition models and four adaptive defenses are proposed to improve the robustness of these systems.
- Effects of Exponential Gaussian Distribution on (Double Sampling) Randomized Smoothing2
An analytic formula is derived for ESG's certified radius, which converges to the origin formula of RS as the dimension $d$ increases, and it is proved that EGG can provide tighter constant factors than DSRS in providing $\Omega(\sqrt{d})$ lower bounds of $\ell_2$ certified radius, and thus further addresses the curse of dimensionality in RS.
- LocalStyleFool: Regional Video Style Transfer Attack Using Segment Anything Model2
LocalStyleFool is proposed, an improved black-box video adversarial attack that superimposes regional style-transfer-based perturbations on videos that can improve both intra-frame and inter-frame naturalness through a human-assessed survey, while maintaining competitive fooling rate and query efficiency.
- Flow-Attention-based Spatio-Temporal Aggregation Network for 3D Mask Detection2
This work proposes a novel 3D mask detection framework, called FASTEN (Flow-Attention-based Spatio-Temporal aggrEgation Network), which can eliminate redundant spatio-temporal feature interference and quickly capture splicing traces of 3D masks in fewer frames.
- 1
- –
- –
- Application of low-altitude sensing technology in low-altitude intelligent networking: a bibliometric study from 2015 to 2025–
Results reveal that UAV-mounted LiDAR and radar systems dominate environmental monitoring—especially forest biomass estimation—and that research has progressed from sensor integration to algorithm optimization and, most recently, AI-driven multimodal sensing.
- TELSAFE: Security Gap Quantitative Risk Assessment Framework–
A new hybrid risk assessment framework called TELSAFE, which employs probabilistic modeling for quantitative risk assessment and eliminates the influence of expert opinion bias is introduced, facilitating effective risk management strategies tailored to the unique requirements of organizations.
- –
- –
- The Improved Conformable Fractional Grey Model Using ADMM and HOA Algorithm–
A fractional calculus prediction model is introduced, utilizing Conformable Fractional Difference and Conformable Fractional Accumulation to calculate the new time series and demonstrates greater accuracy than the original model when dealing with small sample data.
- –
- –
- LAMPS '25: ACM CCS Workshop on Large AI Systems and Models with Privacy and Security Analysis–
This workshop will delve into the complex privacy and cybersecurity risks posed by LAMs in critical sectors, emphasizing the crucial balance between advancing AI technologies and securing critical digital and physical infrastructures.
- –
- –
- Double-Flow-based Steganography without Embedding for Image-to-Image Hiding–
DF-SWE is the first SWE method that can hide large images and multiple images into one image with the same size, significantly enhancing the payload capacity and can be applied in the steganography of secret images in various domains without requiring training data from the corresponding domains.
- –
- Visual Subtitle Feature Enhanced Video Outline Generation–
A Visual Subtitle feature Enhanced video outline generation model (VSENet) is proposed which takes as input the textual subtitles together with their visual font sizes and positions which largely outperforms other baseline methods.
- –
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.