A Quantum LSTM-based approach to cyber threat detection in virtual environment

Quantum information processing (QIP) offers a substantial speed advantage over classical processing, which is particularly promising in the fields of quantum artificial intelligence and quantum machine learning (QAI/QML). This study focuses on addressing the challenge of threat identification in virtualized systems by analyzing system call sequences of malware using quantum long short-term memory (QLSTM) networks. We introduce our dataset along with a straightforward data encoding methodology to prepare the data for variational quantum circuits (VQC), which form the core of the QLSTM models. Additionally, we propose an efficient ansatz tailored to this problem domain, optimizing the QLSTM’s ability to process and analyze complex sequential data effectively. Our research takes a deep dive into the performance of the QLSTM model across various circuit depths, aiming to determine the optimal number of circuit layers relative to the number of qubits utilized. By evaluating the effectiveness of different configurations, we identify the most efficient structure for the QLSTM for our problem domain. This study highlights the importance of finding the right balance between circuit depths and qubit counts and provides insights into the development of efficient quantum machine learning models for cybersecurity applications.