A BI-OBJECTIVE HYPER-HEURISTIC SUPPORT VECTOR MACHINES FOR BIG DATA CYBER-SECURITY
Keywords:
The SVM configuration search, space is efficiently explored, low-level heuristicsAbstract
Researchers have a formidable task in
addressing cyber security concerns in the context
of big data. It has been proposed that machine
learning techniques may be used to deal with
large data security issues. Support vector
machines (SVMs) are one such approach that has
shown exceptional performance on several
classification issues. However, in order to set up
an efficient SVM, the user must first determine
the optimal SVM configuration; this is a difficult
operation that calls for specialised expertise and
much trial and error. In this research, we provide
a formalisation of the SVM configuration
procedure as a bi-objective optimisation problem,
where accuracy and model complexity are treated
as competing goals. We present a new problemdomain-agnostic hyper-heuristic framework for
bi-objective optimisation. For the first time, a
hyper-heuristic has been created specifically for
this issue. High-level strategy and low-level
heuristics are what make up the suggested hyperheuristic framework. In order to determine which
of many possible low-level heuristics should be
utilised to produce a new SVM configuration, the
high-level approach monitors search
performance. The SVM configuration search
space is efficiently explored by the low-level
heuristics, each of which follows a unique set of
rules. The proposed framework adaptively
merges the benefits of decomposition- and
Pareto-based techniques to approximate the
Pareto set of SVM configurations, making it
suitable for bi-objective optimisation. The
suggested methodology has been tested on two
cyber security issues: classifying Microsoft
malware using a large data set and detecting
anomalies in network traffic. In comparison to its
contemporaries and other algorithms, the
acquired results show that the suggested
framework is very successful
Downloads
Published
Issue
Section
License

This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.













