Yihao Wang

ORCID: 0000-0003-2983-8491
Publications
Citations
Views
---
Saved
---
About
Contact & Profiles
Research Areas
  • Advanced Steganography and Watermarking Techniques
  • Internet Traffic Analysis and Secure E-voting
  • Digital Media Forensic Detection
  • Hate Speech and Cyberbullying Detection
  • Human Pose and Action Recognition
  • Anomaly Detection Techniques and Applications
  • Handwritten Text Recognition Techniques
  • Gait Recognition and Analysis
  • Authorship Attribution and Profiling
  • Chaos-based Image/Signal Encryption
  • Spam and Phishing Detection
  • Low-power high-performance VLSI design
  • Infrared Target Detection Methodologies
  • Legal Language and Interpretation
  • Atmospheric Ozone and Climate
  • Vehicle License Plate Recognition
  • Advanced Multi-Objective Optimization Algorithms
  • Music and Audio Processing
  • Evolutionary Algorithms and Applications
  • Remote Sensing and Land Use
  • Rough Sets and Fuzzy Logic
  • Multimodal Machine Learning Applications
  • Metaheuristic Optimization Algorithms Research
  • Speech Recognition and Synthesis
  • Video Surveillance and Tracking Methods

Beijing University of Posts and Telecommunications
2023-2025

University of Chinese Academy of Sciences
2022

Xi'an Institute of Optics and Precision Mechanics
2022

Hong Kong Polytechnic University
2021

Shanghai Maritime University
2021

Henan Normal University
2020-2021

With the widespread use of social media, user-generated content has surged on online platforms. When such includes hateful, abusive, offensive, or cyberbullying behavior, it is classified as toxic speech, posing a significant threat to ecosystem's integrity and safety. While manual moderation still prevalent, overwhelming volume psychological strain human moderators underscore need for automated speech detection. Previously proposed detection methods often rely large annotated datasets;...

10.48550/arxiv.2501.00907 preprint EN arXiv (Cornell University) 2025-01-01

Intelligent transportation systems (ITS) use advanced technologies such as artificial intelligence to significantly improve traffic flow management efficiency, and promote the intelligent development of industry. However, if data in ITS is attacked, tampering or forgery, it will endanger public safety cause social losses. Therefore, this paper proposes a watermarking that can verify integrity copyright response needs ITS, termed ITSmark. ITSmark focuses on functions extracting watermarks,...

10.1109/tits.2025.3535932 article EN IEEE Transactions on Intelligent Transportation Systems 2025-01-01

When the data undergo a distribution change, existing linguistic steganalysis often struggles to effectively capture statistical characteristics of transformed cover or stego, resulting in drop performances. To address this issue and fully use information from original before letter proposes reinforcement learning-based method for steganalysis. This employs an agent (steganalyzer) interact within observation space, enabling adaptation capturing features. Specifically, we map texts GloVe...

10.1109/lsp.2023.3310380 article EN IEEE Signal Processing Letters 2023-01-01

Redundant steganalysis feature components in high-dimensional of images increase the spatio-temporal complexity and even reduce detection accuracy stego images. In order to image dimension, improve achieve fast selection, this paper proposes a general method for selection. Firstly, metric algorithm based on difference function is given, measures between cover class class, which provides basis selecting contributing greatly detect Secondly, Pearson correlation coefficient improved used...

10.1109/access.2020.3018709 article EN cc-by IEEE Access 2020-01-01

Linguistic Steganography (LS) tasks aim to generate steganographic texts (stego) based on secret information. Only authorized recipients can perceive the existence of information in and accurately extract it, thereby preserving privacy. However, controllability stego generated by existing schemes is poor, difficult contain specific discourse characteristics such as style, genre, theme. As a result, are often easily detectable, compromising covert communication. To address these problems,...

10.48550/arxiv.2401.15656 preprint EN arXiv (Cornell University) 2024-01-28

Linguistic steganalysis (LS) tasks aim to effectively detect stegos generated by linguistic steganography. Existing LS methods overlook the distinctive user characteristics, leading weak performance in social networks. The limited occurrence of further complicates detection. In this paper, we propose UP4LS, a novel framework with User Profile for enhancing performance. Specifically, delving into post content, explore attributes like writing habits, psychological states, and focal areas,...

10.48550/arxiv.2311.01775 preprint EN cc-by arXiv (Cornell University) 2023-01-01

To realize the long-period and automated data collection of lunar radiation eliminate geometric errors hyperspectral image during observation process, this paper proposes a slit-type spectrometer method based on rotating table equatorial mount. This uses wide field finder-scope to automatic moon tracking positioning; at same time, it corrects drift angle oversampling collected raw data. The results show that full situation is as non-full situation. can effectively find track position moon's...

10.1117/12.2631853 article EN 2022-04-14

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

10.2139/ssrn.4777618 preprint EN 2024-01-01

To detect stego (steganographic text) in complex scenarios, linguistic steganalysis (LS) with various motivations has been proposed and achieved excellent performance. However, the development of generative steganography, some stegos have strong concealment, especially after emergence LLMs-based existing LS low detection or even cannot them. We designed a novel two modes called LSGC. In generation mode, we created an LS-task "description" used ability LLM to explain whether texts be detected...

10.48550/arxiv.2406.04218 preprint EN arXiv (Cornell University) 2024-06-06

With the evolution of generative linguistic steganography techniques, conventional steganalysis falls short in robustly quantifying alterations induced by steganography, thereby complicating detection. Consequently, research paradigm has pivoted towards deep-learning-based steganalysis. This study offers a comprehensive review existing contributions and evaluates prevailing developmental trajectories. Specifically, we first provided formalized exposition general formulas for steganalysis,...

10.48550/arxiv.2409.01780 preprint EN arXiv (Cornell University) 2024-09-03

Existing temporal action detection algorithms cannot distinguish complete and incomplete actions while this property is essential in many applications. To tackle challenge, we proposed the progression networks (APN), a novel model that predicts of video frames with continuous numbers. Using sequence test video, on top APN, searching algorithm (CAS) was designed to detect only. With usage frame-level fine-grained structure modeling detecting according their whole context, our framework can...

10.1109/icpr48806.2021.9412081 article EN 2022 26th International Conference on Pattern Recognition (ICPR) 2021-01-10

Fewer contribution feature components in the image high-dimensional steganalysis are able to increase spatio-temporal complexity of detecting stego images, and even reduce detection accuracy. In order maintain or improve accuracy while effectively reducing dimension DCTR feature, this paper proposes a new selection approach for feature. First, asymmetric distortion factor information gain ratio each component improved measure difference between symmetric cover features, which provides...

10.3390/sym13101775 article EN Symmetry 2021-09-24
Coming Soon ...