11.07.2015 Views

syssec_red_book

syssec_red_book

syssec_red_book

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

5. Social Networksdisasters. This real-time information is expanding out of all proportion to ourability to process it. This is not just a data mining problem; rather, processingdata streams for identification of malicious content and fraudulent sources isthe main challenge.There have been numerous studies of graph mining, which has led to avariety of techniques for mining relational data. These techniques can beused to characterize the structural properties of social graphs that correspondto normal social behavior. This in turn can lead to methods for detectinganomalies, such as possible attacks, that do not conform to the expected socialbehavior. Although studies of snapshots of social networks can be used forsecurity research, tracking changes over time can give a better insight into thedynamic nature of the network and attackers’ behavior. Therefore, coping withthe dynamicity of data is another big challenge.In studies of social interactions between people, there is a trade-off betweensecurity and privacy. Data collection and the processing of user activity logscan lead to detection of compromised or malicious accounts in social networks;however, this has to be done in a privacy-preserving manner, otherwise it isnot practical.5.6 Example ProblemsWith respect to the above research gaps, the following example problems canbe defined.How can we arrive at a measure of confidence in the truthfulness of informationthat is disseminated through social networks? This is a challengingproblem, particularly in cases where the true identity and the trustworthinessof the source of information are not known, due to anonymity,and in cases where there are no other ways to verify the content.How can we collectively study the information collected from different sourcesin real time? There is a need for an engine for organizing real-time streamingdata gathe<strong>red</strong> from a variety of social sensing platforms, includingsocial networks [388]. How can we effectively parallelize and distributethe data stream processing and introduce methods for identifying cybercriminals based on the aggregated data?How can we utilize data mining techniques for discriminating between honestand malicious identities? The well-studied techniques for graph miningcan be deployed as a tool for combating cyber criminals. Although thestructural properties of social graphs have already been used againstSybil attacks and spam, much more can be done.40

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!