Spam Filtering Security Evaluation Using MILR Classifier
Kunjali Pawar, Madhuri Patil · Automation and Autonomous Systems · 2016
Statistical spam filters are vulnerable to the adversarial attacks. An e-mail is classed as spam if a minimum of one instance within the corresponding bag is spam, and as legitimate if all the instances in its square measure legitimate. These systems based on the design methods and classical methods which do not take into account adversarial settings. In this paper, the security evaluation framework is proposed to avoid the detection in the Spam filtering with the help of Multiple Instance Logistic Regression i.e. MILR. In addition to define the model of Adversary with the guidelines for simulating attack scenarios The principal theme of the framework is to develop an enhanced model which anticipates the attacks by utilizing a data distribution.