題  目：Security of Deep - Battle on the Face

内容簡介：During past several years we have seen many improvements on machine learning technologies which have also been deployed in many real world systems. However, this has also brought some new security problems at the same time. One example is the security of face based authentication system. It has been reported that current face authentication systems can be easily cheated with masks and/or computer generated videos. In the first part of this talk, I will introduce one of my works on performing liveness detections which will check if a given face image is coming from a live person or not. Another  problem is adversarial machine learning, where attackers could craft and feed a ambiguous image into a vulnerable machine learning system, resulting a wrong classification results, meanwhile the problem should not exist for human beings. In this talk, I will introduce a work on how to find malicious image samples for a given machine learning system efficiently and at the same time without being noticed. For example, it was said that machine learning algorithms now can have better. In the last part of this talk, I will introduce a work on data set poisonous problem, i.e., a robust detection algorithm for back-door contaminations inside atraining dataset.

報告人：香港中文大學  張克環  副教授

報告人簡介：主要研究方向是系統與網絡安全，包括移動安全、物聯網安全、Web安全，以及各種側信道攻擊與應用等。2012年畢業于美國印第安納大學布魯明頓分校獲得博士學位。張教授具有豐富的系統開發和系統安全研究經驗，在研究和指導學生方面成果豐碩。赴美攻讀博士學位之前曾在港灣網絡等公司工作，使用Verilog語言開發基于FPGA芯片的骨幹網路由器以及某航天器顯示模塊。研究方面，在系統安全領域的四大頂級會議上發表論文近20篇，主持承擔了多項香港政府和國家自科技基金面上項目，并與國内外的安全領域内的知名專家和高校有合作關系（清華、中科院信工所、美國加州大學Irvine分校、美國俄亥俄州立大學、西安交通大學等）。目前研究團隊在讀博士研究生8人，已指導博士畢業4人（去向阿裡巴巴、山東大學、複旦大學、國防科技大學）。

時  間：2019年12月23日（周一）下午3：30始

地  點：南海樓224室

熱烈歡迎廣大師生參加!

太阳集团1088vip

2019年12月20日