Browser Fingerprinting Identification Using Incremental Clustering Algorithm Based on Autoencoder
Futai Zou, Haochen Zhai · 2021
Browser fingerprinting technology is of great significance in the location of malicious traffic, and it also plays an important role in accurate advertising. The concealment or tampering of the explicit identifier leads to the failure of traditional fingerprinting identification technology. In this paper, we implement browser fingerprinting identification technology based on multiple implicit identifiers which are easily accessible. Especially, we designed an incremental clustering algorithm based on autoencoder. The new data point is treated as an incremental process for the clustering algorithm. Different from traditional clustering algorithms, the newcoming data point is compressed by autoencoder, and then treated as an incremental process so that we just need to compare the distance to the nearest cluster center with the distance threshold for fingerprinting identification, instead of traversing the entire dataset. Experimental results show that our incremental clustering algorithm has distinct advantages, especially in time consumption when comparing with K-means and DBSCAN algorithms, and prove that it is a lightweight algorithm for fingerprinting identification to adapt to the dynamic real environment and has better application prospects.