PROJECT TITLE :

Bits Learning: User-Adjustable Privacy Versus Accuracy in Internet Traffic Classification

ABSTRACT:

Throughout the past decade, a great number of Machine Learning (ML)-based mostly methods have been studied for correct traffic classification. Flow options like the discretizations of the primary 5 packet sizes (PS) and flow ports (FP) are thought-about the best discriminators for per-flow classification. For the primary time, this letter proposes to treat the first n-bits of a flow (BitFlow) as features and compares its overall performance with the well-known ACAS (automated construction of application signatures) that takes the primary n-bytes of a flow (ByteFlow) as features. The results show that BitFlow achieves not only a better classification accuracy but conjointly 1–three orders of magnitude faster speed than ACAS in coaching and classifying. A lot of importantly, this letter additionally proposes to treat the primary n-bits of each of the primary few packet payloads (BitPack) as features, that permits a user-adjustable tradeoff between user privacy protection and classification accuracy maximization. The experiments show that BitPack will significantly outperform BitFlow, PS, and FP.