Compressed features and representations for network traffi c analysis incentralised and edge internet architectures

The COMPACT project aims at creating a new set of tools and methodologies for decreasingthe resource complexity and costs associated with traffi c analysis. The key observation isthat modern traffi c analysis systems are generally based on a feature-based representationof traffi c, rather than on the raw packet-based representation. According to this approach,statistical features are extracted from the captured traffi c and passed as input to machinelearning algorithms trained for the specifi c analysis task to accomplish

The COMPACT project aims at creating a new set of tools and methodologies for decreasingthe resource complexity and costs associated with traffi c analysis. The key observation isthat modern traffi c analysis systems are generally based on a feature-based representationof traffi c, rather than on the raw packet-based representation. According to this approach,statistical features are extracted from the captured traffi c and passed as input to machinelearning algorithms trained for the specifi c analysis task to accomplish