基于多分形分析和主动学习反馈算法的图片垃圾邮件过滤  

On Filtering Image Spam Based on Multifractal Analysis and Active Learning Feedback Algorithm

作　　者：周扬玲[1] 钟剑[2] 邓维[2] 

机构地区：[1]四川中医药高等专科学校,四川绵阳621000 [2]西南大学信息中心,重庆400715

出　　处：《西南师范大学学报（自然科学版）》2013年第10期155-160,共6页Journal of Southwest China Normal University(Natural Science Edition)

基　　金：四川省科技计划基金项目(10JJ8086);中央高校基本科研业务费专项基金项目(XDJK2013C005)

摘　　要：图片垃圾邮件通常由随机变形技术制作,人眼认为内容相同但计算机认为不同,导致常规反垃圾邮件技术无法阻止这类垃圾邮件.根据图片垃圾邮件必然具有相似性、大量性和可变性的特点,提出了一种综合多向小波金字塔多分形分析算法和主动学习反馈驱动半监督支持向量机算法的创新图片垃圾邮件过滤方法.实验结果表明,该方法容易与常规反垃圾邮件系统相结合,而且该方法的效率高、准确性好、假阳性低,通过重复训练,可进一步提高准确性、降低假阳性,适合用于对抗性学习和垃圾邮件过滤.Image spam has usually been produced by a variety of images randomized deformation algo- rithms. The spam can make the message fully legible by the human eye but undistinguishable by the com- mon anti-spare engines. In this paper we have proposed a novel image span： recognition composite method which is a hybrid algorithm of multifractal analysis in multi-orientation wavelet pyramid algorithm and ac tive learning feedback-driven semi-supervised support vector machine algorithm based on three natures of image spam： large quantity, similarity and variability. The experimental results demonstrate that our method is easy to plug into conventional anti-spam system with high efficiency, high accuracy and low false positive rate. The accuracy will be improved and the false positive rate reduced along with more and more retraining. So, the method is fit especially for an adversarial learning and processing like spam filtering.

关 键 词：图片垃圾邮件 多分形分析 主动学习聚类 反馈驱动半监督支持向量机 

分 类 号：TP393.04[自动化与计算机技术—计算机应用技术]