Detecting Image Spam Using Local Invariant Features and Pyramid Match Kernel

Free registration required

Executive Summary

Image spam is a new obfuscating method which spammers invented to more effectively bypass conventional text based spam filters. This paper extracts local invariant features of images and runs a one-class SVM classifier which uses the pyramid match kernel as the kernel function to detect image spam. Experimental results demonstrate that their algorithm is effective for fighting image spam.

  • Format: PDF
  • Size: 655 KB