MODEL EVALUATION FOR EVASIVESMS SPAM DETECTION
Keywords:
Machine Learning, SMS Spam Detection, Evasive Spam, Text Classification, Naïve Bayes, Support Vector Machines, Decision Trees, Deep Learning, Feature Engineering, Spam FilteringAbstract
This paper investigates different machine learning methods for identifying deceptive SMS spam. It is quite difficult to detect evasive spam communications because they use obfuscation to bypass typical filters. Some of the models that were assessed were Deep Learning, Naïve Bayes, Decision Trees, and Support Vector Machines. Included in the compilation are preprocessed messages that are either spam or ham originating from real sources. When comparing results, F1-score, recall, accuracy, and precision are given as appraisal metrics. The results of the experiment highlight the advantages and disadvantages of each paradigm. Deep learning models are superior to more conventional methods when dealing with patterns of varying complexity. To make detection better, we need to augment the data and develop the features. Several recommendations for enhancing spam detection models are included in the article. Modifications to the model will be implemented in the future to address the evolving tactics employed by spammers.
