新任助教講演会(Lectures from New Assistant Professors)

新任助教講演会(Lectures from New Assistant Professors)

日時：	平成28年4月18日(月)3限 (13:30 -- 15:00), 2016/04/18, Monday
場所(Location)：	L1
司会(Chair)：	市川昊平 (Kohei ICHIKAWA)

講演者(Presenter)：	Doudou Fall, インターネット工学研究室 (Internet Engineering Lab.)
題目(Title)：	Risk-Adaptive Authorization Mechanism (RAdAM) for Cloud Computing
概要(Abstract)：	Cloud computing provides many advantages for both the cloud service provider and the clients. It is also infamous for being highly dynamic and for having numerous security issues. The dynamicity of cloud computing implies that dynamic security mechanisms are being employed to enforce its security, especially in regards to access decisions. However, this is surprisingly not the case. Static traditional authorization mechanisms are being used in cloud environments, leading to legitimate doubts on their ability to fulfill the security needs of the cloud. I propose a Risk-Adaptive Authorization Mechnanism (RAdAM) for a simple cloud deployment, collaboration in cloud computing and federation in cloud computing. I use a fuzzy inference system to demonstrate the practicability of RAdAM. I complement RAdAM with a Vulnerability Based Authorization Mechanism (VBAM) which is a real-time authorization model based on the average vulnerability scores of the objects present in the cloud. Finally, i demonstrate the usefulness of VBAM in a case featuring OpenStack.

講演者(Presenter)：	佐々木博昭 (Hiroaki SASAKI), 数理情報学研究室 (Mathematical Informatics Lab.)
題目(Title)：	確率密度関数の微分推定とその機械学習問題への応用について Density derivative estimation and its application to machine learning
概要(Abstract)：	確率密度関数を推定することは，機械学習問題を解く上で，最も一般性のあるアプローチの１つである．一方，多くの機械学習問題では，確率密度関数ではなく，その微分が必要な場合がある．そういった問題に対する自然なアプローチは，確率密度関数の推定を実行することなく，直接，その微分を推定することであろう．本発表では，確率密度関数の微分を必要とする機械学習問題について述べ，その微分を直接推定する手法について紹介する．そして，機械学習問題への応用例についても述べたいと思う． Estimating the probability density function of data is one of the most general approach to machine learning problems. However, a certain number of problems requires to estimate the derivatives rather than the probability density function itself. A natural approach to solve these problems is to directly estimate the derivatives without going through density estimation. In this talk, first, machine learning problems that need density derivative estimation are introduced, and then a method to directly estimate density derivatives is shown. Finally, the usefulness of the estimation method is demonstrated through application to machine learning problems.