3 research outputs found
An integrated model to email spam classification using an enhanced grasshopper optimization algorithm to train a multilayer perceptron neural network
No full textEmail is an important communication that the Internet has made available. One of the significance is seen in the great
ease in which immediate transmission of internet data is done during email transmission. This great ease emerges
with a major issue which is the continuous increase in spam emails. Thus, the need for a spam email detector. The
versatility and adaptability of the nature of spam influenced past innovations. However, previous techniques have been
weakened. This study introduces an email detection model that is designed based on use of an improved version of
the grasshopper optimization algorithm to train a Multilayer Perceptron in classifying emails as ham and spam. To
validate the performance of EGOA, executed on the spam email dataset are utilized, then the performance was
relatively compared with popular search algorithms. The implementation demonstrates that EGOA introduces the best
results with high accuracy of up to 96.09%
Spam classification based on supervised learning using grasshopper optimization algorithm and artificial neural network
No full textThe electronic mailing system has in recent years become a timely and convenient way for the exchange of multimedia
messages across the cyberspace and computer networks in the global sphere. This proliferation has prompted most (if
not all) inboxes receiving junk email messages on numerous occasions every day. Due to these surges in spam attacks,
a number of approaches have been proposed to lessen the attacks across the globe significantly. The effect of previous
detection techniques has been weakened due to the adaptive nature of unsolicited email spam. Hence, resolving spam
detection (SD) problem is a challenging task. A regular class of the Artificial Neural Network (ANN) called Multi-Layer
Perceptron (MLP) was proposed in this study for email SD. The main idea of this research is to train a neural network
by leveraging a new nature-inspired metaheuristic algorithm referred to as a Grasshopper Optimization Algorithm
(GOA) to categorize emails as ham and spam. Evaluation of its performance was performed on an often-used standard
dataset. The results showed that the proposed MLP model trained by GOA achieves high accuracy of up to 94.25%
performance compared to other optimization
Metaheuristic based ids using multi-objective wrapper feature selection and neural network classification
No full textDue to the significant ongoing expansion of computer networks in our lives nowadays, the demand for network
security and protection from cyber-attacks has never been more imperative to either clients or businesses alike, which
signifies the key role of cyber intrusion detection systems in network security. This article proposes a cyber-intrusion
detecting system classification with MLP trained by a hybrid metaheuristic algorithm and feature selection based on
multi-objective wrapper method. The classifier, named as HADMLP is trained using a hybridization of the artificial
bee colony along with the dragonfly algorithm. A multi-objective artificial bee colony model which is wrapper-based
is used for selection of feature. Hence, collective name of the proposed technique referred as MO-HADMLP. For
performance evaluation, the proposed method was assessed using ISCX 2012 and KDD CUP 99 datasets. The results
of our experiments indicate a significant enhancement to the efficacy of network intrusion detection when compared
to other approaches
