Automatic identification and removal of low quality online information

摘要: The advent of the Internet has generated a proliferation online information-rich environments, which provide information consumers with an unprecedented amount freely available information. However, openness these environments also made them vulnerable to new class attacks called Denial Information (DoI) attacks. Attackers launch by deliberately inserting low quality into promote that or deny access high These directly threaten usefulness and dependability as result, important research question is how automatically identify remove this from environments. first contribution thesis set techniques for recognizing countering various forms DoI in email systems. We develop attack based on camouflaged messages, we show spam producers are entrenched arms race. To break free race, propose two solutions. One solution involves refining statistical learning process associating disproportionate weights legitimate features, other leverages existence non-textual features (e.g., URLs) make classification more resilient against second framework collecting, analyzing, classifying examples World Wide Web. fully automatic Web collection technique use it create Webb Spam Corpus—a first-of-its-kind, large-scale, publicly data set. Then, perform large-scale characterization using content HTTP session analysis. Next, present lightweight, predictive approach relies exclusively final detect help prevent within social First, detailed descriptions each novel capturing spam, our collected spammers their behaviors.