Cybercrime: Identification and Prediction Using Machine Learning Techniques.

Area of Science:

• Computer Science
• Cybersecurity
• Data Mining

Background:

• Cybercrime is a growing threat in the digital age.
• Effective cybercrime detection relies on advanced classification and clustering techniques.
• Existing methods require continuous improvement to combat evolving criminal tactics.

Purpose of the Study:

• To evaluate and compare the performance of supervised and unsupervised machine learning models for cybercrime detection.
• To identify the most effective technique for identifying cybercriminals using personal characteristic data.
• To assess the accuracy and various performance metrics of different algorithms.

Main Methods:

• Utilized supervised learning models: Support Vector Machine (SVM) and K-Nearest Neighbor (KNN).
• Employed unsupervised learning techniques: K-means clustering, Gaussian mixture model, and fuzzy clustering methods.
• Applied the Expectation-Maximization (EM) algorithm for Gaussian mixture model performance assessment.
• Analyzed a dataset of 1000 user identities with personal characteristics from CBS open data StatLine.

Main Results:

• The Support Vector Machine (SVM) classifier achieved the highest accuracy at 89% in the supervised method.
• The Gaussian mixture model, an unsupervised method, demonstrated enhanced performance with 76.56% accuracy in criminal detection.
• Comprehensive performance metrics including True Positive (TP), False Positive (FP), accuracy, precision, and recall were computed.

Conclusions:

• Supervised learning, particularly SVM, shows superior performance for cybercrime classification compared to unsupervised methods in this study.
• The Gaussian mixture model offers a viable unsupervised approach for cybercrime detection, with potential for further optimization.
• Accurate cybercrime identification is crucial for law enforcement and cybersecurity, necessitating the use of robust machine learning models.