CIF: SMALL: RUI: Novel Detection Approaches with Comprehensive Hybrid Intelligent Systems for Multimedia Forensics

CIF：SMALL：RUI：用于多媒体取证的综合混合智能系统的新颖检测方法

• 批准号: 1318688
• 负责人: Qingzhong Liu
• 金额: $ 25万
• 依托单位: Sam Houston State University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2014
• 资助国家: 美国
• 起止时间: 2014-01-01 至 2018-08-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1318688&HistoricalAwards=false
• 关键词: CIF; SMALL; RUI; Novel; Detection

Multimedia forensics is increasingly becoming an important research area for protecting public safety and enhancing national security; steganalysis and forgery detection are becoming two active research areas, even as they are still in their inception. The shortage of comprehensive detection systems has severely impeded progress and practical applications in the field. This study will facilitate the work of forensic practitioners in criminal investigations and will deliver to the community benchmark datasets, performance evaluation standards and models, software design and the analysis toolkit in order to design and test new steganalysis/forgery detection system approaches. Additionally, as part of the research education plan, undergraduate women, minorities, and first-generation college students will be extensively involved. A further contribution to an educational pipeline for computer science involves local high school students through the creation of a multimedia forensics and knowledge discovery laboratory. The research involves designing comprehensive hybrid intelligent systems for steganalysis and forgery detection with the capability to discriminate the operations of steganography, general processing and forgery manipulation. This research will make contributions to engineering by providing the optimal classification/regression models associated with the identification of minimal optimal feature sets, benefitting a broad range of applications. The scientific objectives of the research involve validation of new paradigms, shift-recompression-based derivative/wavelet differential pattern analyses on multimedia data, to merge high dimensional rich models that were used for steganalysis, and apply judicious feature mining and machine learning techniques to handle high dimensionality and establish comprehensive hybrid intelligent forensics systems to recognize different categories of multimedia data and reveal past processing history.

多媒体取证正日益成为保护公共安全和加强国家安全的重要研究领域，隐写分析和伪造检测正成为两个活跃的研究领域，尽管它们还处于起步阶段。缺乏全面的检测系统，严重阻碍了该领域的进展和实际应用。这项研究将促进法医从业人员在刑事调查中的工作，并将向社区提供基准数据集、业绩评价标准和模型、软件设计和分析工具包，以设计和测试新的隐写分析/伪造检测系统方法。此外，作为研究教育计划的一部分，本科生女性、少数民族和第一代大学生将广泛参与。通过创建多媒体取证和知识发现实验室，对计算机科学教育渠道的另一项贡献涉及当地高中生。这项研究涉及设计全面的混合智能系统，用于隐写分析和伪造检测，具有识别隐写操作、一般处理和伪造操作的能力。这项研究将通过提供与最小最优特征集识别相关的最优分类/回归模型来为工程做出贡献，从而受益于广泛的应用。这项研究的科学目标包括验证新的范式，对多媒体数据进行基于移位再压缩的导数/小波差分模式分析，合并用于隐写分析的高维丰富模型，并应用明智的特征挖掘和机器学习技术处理高维数据，建立全面的混合智能取证系统，以识别不同类别的多媒体数据并揭示过去的处理历史。