Conceptual

Graph-Embedding and Neural-Network Pipeline for SELinux Policy Anomaly Detection

An automated SELinux security-policy analysis method that represents policies as a graph whose nodes are security contexts and whose edges are access relationships, learns continuous node embeddings with Node2vec, and feeds them to Random Forest, SVM, and multi-layer-perceptron classifiers to flag policy violations. On synthetic datasets derived from production systems the MLP reached about 95% accuracy with balanced precision and recall, detecting separation-of-duty violations, domain-transition issues, and unauthorized access paths.