ANALYSIS OF IDENTIFICATION OF CYBERCRIMES USING CYBER SECURITY ANALYTICS POWERED BY ARTIFICIAL INTELLIGENCE

Cyber-attack attempts to breach the security of computer networks, systems, or data they can use different ways i.e. denial-of-service attack, ransomware, phishing, and hacking with the help of artificial intelligence. Individuals, organizations, or even nation-states may conduct these assaults, which can have dire repercussions such as monetary loss, data breaches, and the interruption of vital infrastructure. Similarly, NotPetya ransomware outbreak was responsible for the shutdown of millions of machines in a span of less than ten minutes. This study investigated the detection of cybercrimes through the use of cyber security analytics driven by artificial intelligence. Additionally, the study intends to uncover hidden patterns and correlations in botnet activities that may not be obvious through the use of traditional methods. Increasing the intelligence gathered about cyber threats and making it easier to take preventative forensic steps are both extremely important. Within the context of Noakhali Science and Technology University in Bangladesh 2024, the experimental study was carried for analyzing and identifying cybercrimes through the utilization of cyber security analytics driven by artificial intelligence. For the development of our method, we utilized the Python programming language in conjunction with the Google Colab and TensorFlow frameworks. Python's selection was based on its many positive characteristics, which include its minimal code requirements, vast availability of libraries and frameworks, consistency, independence from platform, a flourishing community, and adaptability. The model employs sophisticated artificial intelligence techniques in order to recognize botnets as a key source of cyberattacks. Studying cybersecurity and botnet identification required the usage of the CTU-13 dataset as well as the IoT-23 dataset. In order to prepare the data, we employed CNN, LSTM, and Dropout layers, as well as loaded sixteen datasets into a Pandas database and labeled them. The datasets were divided into training and testing sets, and consistent approaches were utilized in order to maintain the training values from the training sets. The results of the tests not only emphasize the extraordinary performance of LSTM in the particular context of the feature hybrid phase, but they also strongly imply that CNN played a key role in strengthening the overall efficacy of the algorithm for the purpose of improving its overall effectiveness. Moreover, the accuracy of botnet identification in the Internet of Things is greater than 92%. Obtaining a very acceptable success rate (around 98.7%) and a false positive rate of 0.04%, the testing findings reveal that the suggested method achieves exceptionally high levels of performance.