MalNet-Quantum: a CNN-based malware detection framework enhanced by Quantum Grey Wolf Optimizer

The Android malware is one of the persistent threats that cannot be ignored, as the evolution of the family is fast, and the distribution of apps is deep and exists outside of the official platforms. Traditional detectors, both signature-based and ML-driven, tend to rely on brittle feature engineering and ad-hoc hyperparameter tuning, limiting their accuracy and damaging their generalization. The current paper introduces MalNet-Quantum, a detector built on CNN where all essential hyperparameters (e.g., kernel size, number of filters, dropout, learning rate) are optimized automatically with the help of a Quantum Grey Wolf Optimizer (QGWO). The pipeline is capable of preserving both static feature vectors (permissions, API usage, network indicators) and binary to image representation, which enables the CNN to learn discriminative spatial patterns without designing features manually. QGWO offers wide range, exploration-based search dynamics, which minimize premature convergence, and outperform manual, grid, or random search by a significant margin. Large-scale experiments on benchmark Android malware data indicate that MalNet-Quantum is better than competitive baselines by all evaluation metrics. Specifically, the offered approach has 99.82% accuracy, 99.60% precision, and 99.85% recall, and the results have statistically significant improvement based on Wilcoxon tests (p < 0.05). These findings suggest that CNN feature learning with quantum-inspired hyperparameter optimization offers a solid and scalable path forward in the Android malware detection to minimize false alarms and the number of missed threats at a high cross-dataset reliability cost.