Android BotNet attacks have increased in number due to the widespread use of mobile applications and gaming. Among these attacks, one Remote Access Trojan (RAT) that has gained attention is DroidJack, exemplified by its presence in the Pokémon GO game application. DroidJack is a notorious RAT that poses a significant cybersecurity threat by targeting Android devices, turning them into part of a botnet, and allowing cybercriminals to gain unauthorized access to and remotely control the devices for malicious activities. Numerous studies demonstrate that applying machine learning (ML) methodologies is an effective approach to identifying and preventing botnet attacks. Using the Android Mischief Dataset, our paper presents classification reports utilizing machine learning techniques such as logistic regression, decision tree classifier, K-nearest neighbor, linear support vector classifier (SVC), linear stochastic gradient descent (SGDC), and eXtreme gradient boost classifier (XGBoost). These models detected RAT attacks with a high accuracy of 99.81%. Alternative methods for identifying botnet and RAT attacks using graph neural network methods surpass current techniques in assessment metrics and include a botnet detector and an explanation for automated forensics. Studies show that models detect botnet connections with higher-than-expected accuracy using methods like multilayer perceptron, deep neural network, and generative adversarial network (GAN). These findings demonstrate the value of machine learning methods in identifying malware attack patterns and detecting botnets, thereby strengthening security systems against such dangers.