A novel machine learning approach for detecting first-time-appeared malware

Engineering Applications of Artificial Intelligence, Год журнала: 2024, Номер 131, С. 107801 - 107801

Опубликована: Фев. 2, 2024

Conventional malware detection approaches have the overhead of feature extraction, requirement domain experts, and are time-consuming resource-intensive. Learning-based mainstay as they overcome most these challenges by significantly improving effectiveness providing a low false positive rate. The exponential growth variants first-time-appeared malware, which includes polymorphic zero-day attacks, some significant to learning-based detectors. These catastrophic impacts on This paper proposes novel deep framework detect effectively efficiently better performance than conventional approaches. First, it translates visualises each Windows portable executable (PE) file into coloured image eliminate extraction need for experts analyse features. In subsequent step, fine-tuned learning model is used extract features from last fully connected layer. step has reduced cost training required models if end-to-end classification. third selects important influential through powerful selection algorithm. then fed one-class classifier final detection. With classifier, an enclosed boundary around benign data constructed. Anything outside declared anomaly/malicious. It enhanced framework's ability evolving, unseen, polymorphic, well reducing problem overfitting. proposed validated with state-of-the-art outperformed accuracy 99.30% Malimg dataset. Wilcoxon signed-rank test validate statistical significance framework. evident results that effective can be in defence industry, resulting more robust solutions against attacks.

Язык: Английский

---

Android malware detection and identification frameworks by leveraging the machine and deep learning techniques: A comprehensive review

Telematics and Informatics Reports, Год журнала: 2024, Номер 14, С. 100130 - 100130

Опубликована: Март 12, 2024

The ever-increasing growth of online services and smart connectivity devices have posed the threat malware to computer system, android-based phones, Internet Things (IoT)-based systems. anti-malware software plays an important role in order safeguard system resources, data information against these attacks. Nowadays, writers used advanced techniques like obfuscation, packing, encoding encryption hide malicious activities. Because evasion, traditional detection unable detect new variants malware. Cyber security has attracted many researchers past for designing Machine Learning (ML) or Deep (DL) based models. In this study, we present a comprehensive review literature on approaches. overall is grouped into three categories such as feature selection (FS) proposed detection, ML-based DL-based detection. Based review, identified shortcoming research gaps along with some future directives design efficient identification framework.

Язык: Английский

---

Development of hybrid models based on deep learning and optimized machine learning algorithms for brain tumor Multi-Classification

Expert Systems with Applications, Год журнала: 2023, Номер 238, С. 122159 - 122159

Опубликована: Окт. 18, 2023

Язык: Английский

---

Deep neural network with empirical mode decomposition and Bayesian optimisation for residential load forecasting

Expert Systems with Applications, Год журнала: 2023, Номер 237, С. 121355 - 121355

Опубликована: Сен. 9, 2023

In the context of a resilient energy system, accurate residential load forecasting has become non-trivial requirement for ensuring effective management and planning strategy/policy development. Due to highly stochastic nature profiles, it is difficult predict accurately, usually, predictions are error-prone. This paper explores potential Empirical Mode Decomposition (EMD) in simplifying dynamics complex demand profiles. The simplified components then embedded within deep learning model, specifically Convolution Neural Network (CNN) Long Short-Term Memory (LSTM), forecast short-term loads. novel modelling framework integrates Bayesian optimisation strategy, feature decomposition technique, engineering phase, percentile-based bias correction algorithm enhance model accuracy. developed using case-study dwelling located Fintry (Scotland), performance assessed over four horizons. overall efficiency also investigated three algorithms: random forest, gradient boosting decision trees (GBDT), an LSTM network. While EMD were found greatly improve prediction accuracy, number IMFs used was shown significantly impact model's computational complexity. tested on two further case studies from Fintry.

Язык: Английский

---

Student Cheating Detection in Higher Education by Implementing Machine Learning and LSTM Techniques

Sensors, Год журнала: 2023, Номер 23(8), С. 4149 - 4149

Опубликована: Апрель 20, 2023

Both paper-based and computerized exams have a high level of cheating. It is, therefore, desirable to be able detect cheating accurately. Keeping the academic integrity student evaluations intact is one biggest issues in online education. There substantial possibility dishonesty during final since teachers are not directly monitoring students. We suggest novel method this study for identifying possible exam-cheating incidents using Machine Learning (ML) approaches. The 7WiseUp behavior dataset compiles data from surveys, sensor data, institutional records improve well-being performance. offers information on achievement, attendance, general. In order build models predicting accomplishment, at-risk students, detecting problematic behavior, designed use research Our model approach surpassed all prior three-reference efforts with an accuracy 90% used long short-term memory (LSTM) technique dropout layer, dense layers, optimizer called Adam. Implementing more intricate optimized architecture hyperparameters credited increased accuracy. addition, could been caused by how we cleaned prepared our data. More investigation analysis required determine precise elements that led model's superior

Язык: Английский

---

Automated Android Malware Detection Using Optimal Ensemble Learning Approach for Cybersecurity

IEEE Access, Год журнала: 2023, Номер 11, С. 72509 - 72517

Опубликована: Янв. 1, 2023

Current technological advancement in computer systems has transformed the lives of humans from real to virtual environments. Malware is unnecessary software that often utilized launch cyber-attacks. variants are still evolving by using advanced packing and obfuscation methods. These approaches make malware classification detection more challenging. New techniques different conventional should be for effectively combating new variants. Machine learning (ML) methods ineffective identifying all complex The deep (DL) method can a promising solution detect This paper presents an Automated Android Detection Optimal Ensemble Learning Approach Cybersecurity (AAMD-OELAC) technique. major aim AAMD-OELAC technique lies automated identification malware. To achieve this, performs data preprocessing at preliminary stage. For process, follows ensemble process three ML models, namely Least Square Support Vector (LS-SVM), kernel extreme machine (KELM), Regularized random vector functional link neural network (RRVFLN). Finally, hunter-prey optimization (HPO) approach exploited optimal parameter tuning DL it helps accomplish improved results. denote supremacy method, comprehensive experimental analysis conducted. simulation results portrayed over other existing approaches.

Язык: Английский

---

Advances in Malware Analysis and Detection in Cloud Computing Environments: A Review

International Journal of Safety and Security Engineering, Год журнала: 2024, Номер 14(1), С. 225 - 230

Опубликована: Фев. 29, 2024

Cloud computing, integral for data storage and online services, presents significant advantages over traditional distribution methods, including enhanced convenience, on-demand storage, scalability, cost efficiency.Its growing adoption in securing Internet of Things (IoT) cyber-physical systems (CPS) against various cyber threats offers numerous opportunities.Despite the continuous evolution malware lack a universally effective detection method, cloud environments provide promising approach detection.Cloud recognized its efficiency, flexibility, reliability on elastic resources, is widely utilized IT industry to support infrastructure services.However, one foremost security challenges faced attacks.Certain antivirus scanners struggle detect metamorphic or encrypted due complexity scale, allowing such evade detection.High rates with precision reducing false positives are essential.Machine learning (ML) classifiers, vital component Artificial Intelligence (AI) systems, require training extensive volumes develop credible models high rates.Traditional methods face identifying complex malware, as modern employs contemporary packaging obfuscation techniques circumvent measures.This paper provides detailed discussion detecting computing safeguarding IoT CPS from attacks.It survey analysis models, aiding researchers limitations inspiring design innovative quality service levels.

Язык: Английский

---

An Intelligent Mechanism to Detect Multi-Factor Skin Cancer

Diagnostics, Год журнала: 2024, Номер 14(13), С. 1359 - 1359

Опубликована: Июнь 26, 2024

Deep learning utilizing convolutional neural networks (CNNs) stands out among the state-of-the-art procedures in PC-supported medical findings. The method proposed this paper consists of two key stages. In first stage, deep sequential CNN model preprocesses images to isolate regions interest from skin lesions and extracts features, capturing relevant patterns detecting multiple lesions. second stage incorporates a web tool increase visualization by promising patient health diagnoses. was thoroughly trained, validated, tested database related HAM 10,000 dataset. accomplished an accuracy 96.25% classifying lesions, exhibiting significant areas strength. results achieved with validated evaluation methods user feedback indicate substantial improvement over current for lesion classification (malignant/benign). comparison other models, surpasses transfer (87.9%), VGG 19 (86%), ResNet-50 + VGG-16 (94.14%), Inception v3 (90%), Vision Transformers (RGB images) (92.14%), Entropy-NDOELM (95.7%). findings demonstrate potential learning, networks, disease detection classification, eventually revolutionizing melanoma and, thus, upgrading consideration.

Язык: Английский

---

Cybersecurity in the AI era: analyzing the impact of machine learning on intrusion detection

Knowledge and Information Systems, Год журнала: 2025, Номер unknown

Опубликована: Фев. 19, 2025

Язык: Английский

---

Tackling class imbalance in computer vision: a contemporary review

Artificial Intelligence Review, Год журнала: 2023, Номер 56(S1), С. 1279 - 1335

Опубликована: Июль 20, 2023

Язык: Английский

---