题目: Feature Mining with Machine Intelligence in Multimedia Forensics
报告人： Qingzhong Liu, Associate Professor, Sam Houston State University
报告人简介：Qingzhong Liu is currently a tenured Associate Professor in Computer Science at Sam Houston State University (SHSU). His research interests include multimedia forensics, information assurance, bioinformatics, pattern recognition, computational intelligence and applications.
He obtained his Master degree from Sichuan University in 1997, and Ph.D. from New Mexico Institute of Mining and Technology (New Mexico Tech) in 2007. After completing Ph.D. study, he declined a position of research fellow from Harvard University and Massachusetts General Hospital, and continued staying at New Mexico Tech as a senior research scientist, working on a US congress and military funding-based research project “Computational Analysist of Cyber Terrorism against the U.S.”. He joined SHSU in 2010.
His study in multimedia forensics has been sponsored by the National Institute of Justice and the National Science Foundation. He has been serving on the Technical Program Committee for the primary conference in multimedia forensics, ACM Workshop on Information Hiding and Multimedia Security since 2013. He received the Greater Houston Fraud Impact Award from the Houston Chapter of the Association of Certified Fraud Examiners in 2015, the Faculty Research Excellence Award from the College of Science and Engineering Technology of the Sam Houston State University in 2016, and the Excellence in Scholarly and Creative Accomplishments Award from Sam Houston State University in 2017 and he is the sole CS professor and the third Chinese professor honored with such the highest achievement in the SHSU history.
He has published over 100 peer-reviewed papers. According to the data from google scholar, the citations to his work are over 1700. He is anticipated to serve on the core for the PhD program in Digital and Cyber Forensic Science, the first such a PhD program in U.S.A., which will be launched in the fall 2018 at SHSU.
报告内容: While enormous digital multimedia data nowadays mostly enrich our everyday life, cyber terrorists and cyber criminals may use the data to conduct stealthy communication by steganography, or alter the truth and make fraudulent illusions by forgery manipulation. The threats posed by hackers, spies, terrorists, and criminals, etc. using steganography and advanced multimedia forgery techniques for illegal purposes are realistic for us to concern. To combat and defeat such threats, steganalysis and forgery detection have become two spots in multimedia forensics.
In this talk, the following contents in steganalysis and forgery detection will be covered: (1) integrated with information hiding ratio, the concept of image complexity will be discussed to enhance the evaluation of steganalysis performance; (2) calibrated neighboring joint density-based approaches will be introduced to distinguish steganograms from covers in JPEG images; (3) with the aid of ensemble learning, hybrid large feature mining-based approaches will be presented, in order to address the highly challenging image forensics problems including the detection of JPEG down-recompression, advanced seam-carving and inpainting forgery under JPEG compound anti-forensics; and (4) some open problems in multimedia forensics will be discussed.