RESEARCHERS are advancing efforts to derail online child exploitation by determining which proactive attack strategies are the most effective.
In a new preprint paper, Simon Fraser University researcher Richard Frank and Golestan University researcher Fateme Movahedi found greater efficacy in combating online exploitation using a digital attack strategy known as “principal component analysis,” or PCA. This method is designed to hone in on key nodes and connections in complex networks, and can be applied across a wide range of fields.
“Agencies that combat this problem have limited resources, therefore it is important to access the most efficient attack strategies,” says Frank, an assistant professor in the School of Criminology and director of the International Cybercrime Centre at SFU.
“There are multiple existing strategies for disrupting online child exploitation networks that are dependent on some properties of networks. We sought to determine what would cause the largest disruption to the networks themselves.”
Two PCA algorithms tested, known as the adjacency matrix (PCA_A) and the Laplacian matrix (PCA_L), were found to be more effective in attacking the structure of a single online network of child exploitation websites when compared to other established attack methods, including hubs, which highlight visible breaches; bridges, which can connect network segments; network capital, which measures how people communicate through electronic networks; and random searches.
While the hub attack produced results close to the PCA strategies the latter performed better overall, selecting the nodes that corresponded to websites yielding significant information to the network.
The attack strategies were tested on network data collected from the Internet by a custom-written web crawler. Once PCA identified the key players from the data collected those websites were removed from the study dataset.
Frank, who is also founder of The Dark Crawler, a tool that has collected more than 150 million posts related to online violent extremist activities, says future research should further consider the effectiveness of these attack methods in other types of networks beyond those extracted from the internet.