Search

Data poisoning attacks on machine learning models aim to manipulate the data used for model training such that the trained model behaves in the attacker's favor. In classical models such as deep neural networks, large chains of dot products do indeed cause errors injected by an attacker to propagate or to accumulate. But what about quantum models? Our hypothesis is that, in quantum machine learning, error propagation is limited for two reasons. First of all, data, which in quantum computing is encoded in terms of qubits which are confined to the Bloch sphere. Second of all, quantum information processing happens via the application of unitary operators which are norm-preserving. Testing this hypothesis, we investigate how extensive error propagation and thus poisoning attacks affect quantum machine learning.

Over the last decade, Artificial Intelligence (AI) has become increasingly popular, especially with the use of chatbots such as ChatGPT, Gemini, and DALL-E. With this rise, large language models (LLMs) and Generative AI (GenAI) have also become more prevalent in everyday use. These advancements strengthen cybersecurity's defensive posture and open up new attack avenues for adversaries as well. This paper provides a comprehensive overview of the current state-of-the-art deployments of GenAI, covering assaults, jailbreaking, and applications of prompt injection and reverse psychology. This paper also provides the various applications of GenAI in cybercrimes, such as automated hacking, phishing emails, social engineering, reverse cryptography, creating attack payloads, and creating malware. GenAI can significantly improve the automation of defensive cyber security processes through strategies such as dataset construction, safe code development, threat intelligence, defensive measures, reporting, and cyberattack detection. In this study, we suggest that future research should focus on developing robust ethical norms and innovative defense mechanisms to address the current issues that GenAI creates and to also further encourage an impartial approach to its future application in cybersecurity. Moreover, we underscore the importance of interdisciplinary approaches further to bridge the gap between scientific developments and ethical considerations., Comment: 40 pages

Let $p$ and $q$ be two distinct odd primes, $p

For the first time, we enable the execution of hybrid machine learning methods on real quantum computers with 100 data samples and real-device-based simulations with 5,000 data samples, thereby outperforming the current state of research of Suryotrisongko and Musashi from 2022 who were dealing with 1,000 data samples and quantum simulators (pure software-based emulators) only. Additionally, we beat their reported accuracy of $76.8\%$ by an average accuracy of $91.2\%$, all within a total execution time of 1,687 seconds. We achieve this significant progress through two-step strategy: Firstly, we establish a stable quantum architecture that enables us to execute HQML algorithms on real quantum devices. Secondly, we introduce new hybrid quantum binary classification algorithms based on Hoeffding decision tree algorithms. These algorithms speed up the process via batch-wise execution, reducing the number of shots required on real quantum devices compared to conventional loop-based optimizers. Their incremental nature serves the purpose of online large-scale data streaming for DGA botnet detection, and allows us to apply hybrid quantum machine learning to the field of cybersecurity analytics. We conduct our experiments using the Qiskit library with the Aer quantum simulator, and on three different real quantum devices from Azure Quantum: IonQ, Rigetti, and Quantinuum. This is the first time these tools are combined in this manner., Comment: 32 pages, 4 figures, 5 tables