Traffic classification has received significant attention due to the capability of blocking unwanted transfer of complex information. One of the important decisions that have to be made while constructing a classification model is to employ learning approach. Hierarchical Distributed Peer-to-Peer (HP2PC) architecture grouped to form higher level neighborhoods but elaborated technique were not adopted to handle bidirectional traffic and was not extended to dynamic structure. Cluster-Adaptive Distance Bound (CADB) based on separating hyper plane boundaries of Voronoi clusters facilitate resourceful spatial filtering, with relatively small preprocessing storage. The storage overhead was in a way addressed using the Euclidean and Mahalanobis similarity measures but effective attribute selection was not applied to solve the issue related to distance bounds. To overcome the bidirectional traffic problem with dynamic structure Subjective Symmetrical Support Vector Machine (SS-SVM) mechanism is developed. A hybrid attribute selection algorithm is designed which pre-filters (i.e., Classifies) the attributes with improved (refers to subjective symmetry) Support Vector Machine and solve the related distance bounds problem. After classifying the attributes with SS-SVM, the best attributes are further selected and then generates the attribute value. SS-SVM algorithm assigns higher values to the attributes that are capable to generate data report from minority and majority class. Moreover, SS-SVM for flow-based attribute selection in traffic classification is applied. Subjective Symmetrical mechanism experimented with factors such as classification rate, true positive rate, attribute selection efficiency, memory consumption, and report generation effectiveness. Subjective Symmetrical mechanism improves averagely the attribute selection up to 6% when compared with the state-of-art methods. © 2005 - 2014 JATIT & LLS. All rights reserved.