On Neural Networks and the Future of Spam

Abstract

Content-based filters (e.g. Keyword Filters, Heuristics Filters, Statistical Learning
Filters, Pattern Recognition Neural Networks, and so on) use tokens, which are found during
message content analysis, to separate spam from legitimate messages. The effectiveness of these
token-based filters is due to the presence of token signatures (e.g. tokens that are invariant for the
many variants of spam messages). In our research, we discovered a new trend of spam messages
that have a low frequency of token signatures, thus making them significantly more difficult to
identify. Further on, we will describe this new type of spam and also suggest a few modalities to
combat the spread of this prototype of the future spam trend.