Deep learning-based improved transformer model on android malware detection and classification in internet of vehicles

Scientific Reports, Journal Year: 2024, Volume and Issue: 14(1)

Published: Oct. 24, 2024

With the growing popularity of autonomous vehicles (AVs), confirming their safety has become a significant concern. Vehicle manufacturers have combined Android operating system into AVs to improve consumer comfort. However, diversity and weaknesses pose substantial risks AVs, as these factors can expose them threats, namely malware. The advanced behaviour multi-data source fusion in driving models mitigated recognition accuracy effectualness for To efficiently counter new malware variants, novel techniques distinct from conventional methods must be utilized. Machine learning (ML) cannot detect every complex variant. deep (DL) model is an efficient tool detecting various variants. This manuscript proposes Deep Learning-Based Improved Transformer Model on Malware Detection (DLBITM-AMD) technique Internet (IoVs). main aim presented DLBITM-AMD approach effectually accurately. method performs Z-score normalization process convert raw data standard form. Then, utilizes binary grey wolf optimization (BGWO) select optimum feature subsets. An improved transformer integrated with RNN softmax enhance classification recognition. Finally, snake optimizer algorithm (SOA) employed parameter method. extensive experiment accomplished benchmark dataset. performance validation portrayed superior value 99.26% over existing models.

Language: Английский

---

Earthworm Optimization Algorithm Based Cascade LSTM-GRU Model for Android Malware Detection

Cyber Security and Applications, Journal Year: 2025, Volume and Issue: unknown, P. 100083 - 100083

Published: Jan. 1, 2025

Language: Английский

---

Feature-Driven Malware Detection using Cascade Machine Learning Models

Research Square (Research Square), Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 9, 2025

Malware proliferation continues to jeopardize global data security and user privacy, necessitating robust detection classification mechanisms. In this research, we propose Detection using Cascade Machine Learning (MDCML) classifier designed detect anomalies in Portable Executable (PE) files classify them into malware families with high precision. The model integrates three machine learning algorithms such as Random Forest, Bagging Boosting, fine-tuned through extensive hyperparameter optimization, significantly enhancing performance. To extract features from raw textual data, have utilized a TF-IDF-based inter-class dispersion architecture, transforming unstructured opcode structured feature maps that emphasize contextual importance. employs gradient descent regularization iteratively minimize the loss function prevent overfitting, achieving sublinear regret convergence toward optimal performance.The proposed is validated public Big 2015 dataset, which includes approximately 10,000 spanning nine families. study included comprehensive experimentation on both binary (Malware vs. Benign) multi-class tasks. Performance was evaluated across diverse sample sizes, execution times, optimization strategies ensure analysis. An accuracy of 98.97% highlights superior performance framework over traditional models, showcasing significant advancements. This research underscores concept hybrid MDCML improving classification, thereby privacy.

Language: Английский

---

Android Malware Detection Based on Informative Syscall Subsequences

IEEE Access, Journal Year: 2024, Volume and Issue: unknown, P. 1 - 1

Published: Jan. 1, 2024

The Android operating system commands a dominant market share of over 70% in the smartphone industry.However, this widespread usage has resulted concerning increase malware applications.While existing static detection mechanisms are vulnerable to code obfuscation attacks, manipulating runtime call (syscall) sequence remains significant challenge for attackers.Consequently, syscall-based gaining prominence.Current approaches rely on machine learning algorithms, utilizing numerical features such as syscall frequencies and transition probability matrices.However, wide range values these necessitates large datasets effective classifier training, susceptibility noise outliers persists.As result, there is an urgent need binary representation dynamic improve efficiency.To address challenge, our paper proposes innovative subsequence-based feature method learning-driven detection.By employing information gain method, we identify informative subsequences.The proposed mechanism achieves impressive 99% accuracy detecting applications using just 50% training data, across both Drebin/AMD CICMalDroid2020 datasets.

Language: Английский

---

Application of Deep Learning Models for Real-Time Automatic Malware Detection

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 107742 - 107756

Published: Jan. 1, 2024

The increase in the sophistication and volume of cyberattacks has made traditional malware detection methods, such as those based on signatures heuristics, obsolete. These conventional techniques struggle to identify new variants that employ advanced evasion tactics, resulting significant security gaps. This study addresses this problem by proposing a hybrid model deep learning integrates static dynamic analysis improve precision robustness detection. proposal combines extraction features from code behavior at runtime, using convolutional neural networks for visual recurrent sequential analysis. comprehensive integration allows our detect known more effectively. results show achieves 98%, recall 97%, an F1-score 0.975, outperforming which generally reach 88% 89% precision. Furthermore, outperforms recent approaches documented literature, report up 96% In work, it offers advancement detection, providing effective adaptable solution modern cyber threats.

Language: Английский

---