Hybrid deep learning model for accurate and efficient android malware detection using DBN-GRU DOI Creative Commons
SK. Heena Kauser,

V. Maria Anu

PLoS ONE, Год журнала: 2025, Номер 20(5), С. e0310230 - e0310230

Опубликована: Май 19, 2025

The rapid growth of Android applications has led to an increase in security threats, while traditional detection methods struggle combat advanced malware, such as polymorphic and metamorphic variants. To address these challenges, this study introduces a hybrid deep learning model (DBN-GRU) that integrates Deep Belief Networks (DBN) for static analysis Gated Recurrent Units (GRU) dynamic behavior modeling enhance malware accuracy efficiency. extracts features (permissions, API calls, intent filters) (system network activity, inter-process communication) from APKs, enabling comprehensive application behavior.The proposed was trained tested on the Drebin dataset, which includes 129,013 (5,560 123,453 benign).Performance evaluation against NMLA-AMDCEF, MalVulDroid, LinRegDroid demonstrated DBN-GRU achieved 98.7% accuracy, 98.5% precision, 98.9% recall, AUC 0.99, outperforming conventional models.In addition, it exhibits faster preprocessing, feature extraction, classification times, making suitable real-time deployment.By bridging methodologies, enhances capabilities reducing false positives computational overhead.These findings confirm applicability real-world applications, offering scalable high-performance solution.

Язык: Английский

MINDPRES: A Hybrid Prototype System for Comprehensive Data Protection in the User Layer of the Mobile Cloud DOI Creative Commons
Noah Oghenefego Ogwara, Krassie Petrova, Mee Loong Yang

и другие.

Sensors, Год журнала: 2025, Номер 25(3), С. 670 - 670

Опубликована: Янв. 23, 2025

Mobile cloud computing (MCC) is a technological paradigm for providing services to mobile device (MD) users. A compromised MD may cause harm both its user and other MCC customers. This study explores the use of machine learning (ML) models stochastic methods protection Android MDs connected cloud. To test validity feasibility proposed methods, adopted proof-of-concept approach developed prototype system named MINDPRESS. The static component MINDPRES assesses risk apps installed on MD. It uses device-based ML model feature analysis cloud-based evaluator. hybrid monitors app behavior in real time. deploys two functions as an intrusion detection prevention (IDPS). performance evaluation results showed that accuracy achieved by compared well with reported recent work. Power consumption data indicated did not create overload. contributes feasible scalable framework building distributed systems devices

Язык: Английский

Процитировано

0
FICConvNet: A Privacy-Preserving Framework for Malware Detection Using CKKS Homomorphic Encryption DOI Open Access

Si Pang,

Jing Wen, Sen Liang

и другие.

Electronics, Год журнала: 2025, Номер 14(10), С. 1982 - 1982

Опубликована: Май 13, 2025

Recent advancements in cloud computing, edge and Internet of Things (IoT) have increased the complexity network environments provided fertile ground for malicious attacks. Existing DL-based malware detections, while making progress detection accuracy generalization ability, face serious challenges user data privacy protection. To address this problem, paper proposed a non-interactive system based on CKKS homomorphic encryption (FICConvNet). The effectively achieves end-to-end protection, ensures that sensitive uploaded by users are processed an encrypted state, prevents leakage, protects results. key technology FICConvNet is its innovative lightweight ciphertext inference architecture, which combines DS Conv structured sparse projection to significantly reduce computation. Meanwhile, paper, adaptive learnable activation function (ALPolyAct) designed replace traditional fixed polynomial enhance expressive power model. In addition, protection security results optimized zero-decryption process. Experimental show 95.86%, outperforms existing model CryptoNets (15.5% improvement) approaches performance plaintext ResNet-18. reduces time about 80% compared Conv2d structures. research provides effective privacy-preserving solution field explores new directions application detection.

Язык: Английский

Процитировано

0
Hybrid deep learning model for accurate and efficient android malware detection using DBN-GRU DOI Creative Commons
SK. Heena Kauser,

V. Maria Anu

PLoS ONE, Год журнала: 2025, Номер 20(5), С. e0310230 - e0310230

Опубликована: Май 19, 2025

The rapid growth of Android applications has led to an increase in security threats, while traditional detection methods struggle combat advanced malware, such as polymorphic and metamorphic variants. To address these challenges, this study introduces a hybrid deep learning model (DBN-GRU) that integrates Deep Belief Networks (DBN) for static analysis Gated Recurrent Units (GRU) dynamic behavior modeling enhance malware accuracy efficiency. extracts features (permissions, API calls, intent filters) (system network activity, inter-process communication) from APKs, enabling comprehensive application behavior.The proposed was trained tested on the Drebin dataset, which includes 129,013 (5,560 123,453 benign).Performance evaluation against NMLA-AMDCEF, MalVulDroid, LinRegDroid demonstrated DBN-GRU achieved 98.7% accuracy, 98.5% precision, 98.9% recall, AUC 0.99, outperforming conventional models.In addition, it exhibits faster preprocessing, feature extraction, classification times, making suitable real-time deployment.By bridging methodologies, enhances capabilities reducing false positives computational overhead.These findings confirm applicability real-world applications, offering scalable high-performance solution.

Язык: Английский

Процитировано

0