Biblio

Machine learning is a major area in artificial intelligence, which enables computer to learn itself explicitly without programming. As machine learning is widely used in making decision automatically, attackers have strong intention to manipulate the prediction generated my machine learning model. In this paper we study about the different types of attacks and its countermeasures on machine learning model. By research we found that there are many security threats in various algorithms such as K-nearest-neighbors (KNN) classifier, random forest, AdaBoost, support vector machine (SVM), decision tree, we revisit existing security threads and check what are the possible countermeasures during the training and prediction phase of machine learning model. In machine learning model there are 2 types of attacks that is causative attack which occurs during the training phase and exploratory attack which occurs during the prediction phase, we will also discuss about the countermeasures on machine learning model, the countermeasures are data sanitization, algorithm robustness enhancement, and privacy preserving techniques.

Nowadays, data has become more important as the core resource for the information society. However, with the development of data analysis techniques, the privacy violation such as leakage of sensitive data and personal identification exposure are also increasing. Differential privacy is the technique to satisfy the requirement that any additional information should not be disclosed except information from the database itself. It is well known for protecting the privacy from arbitrary attack. However, recent research argues that there is a several ways to infer sensitive information from data although the differential privacy is applied. One of this inference method is to use the correlation between the data. In this paper, we investigate the new privacy threats using attribute correlation which are not covered by traditional studies and propose a privacy preserving technique that configures the differential privacy's noise parameter to solve this new threat. In the experiment, we show the weaknesses of traditional differential privacy method and validate that the proposed noise parameter configuration method provide a sufficient privacy protection and maintain an accuracy of data utility.