Convincing Features
Assignment Type
Subject
Uploaded by Malaysia Assignment Help
Date
Annotated Bibliography
The rise of mobile devices has led to increased accessibility to information, but it also brings vulnerability. Malicious apps can exploit unsolicited permissions to access sensitive user data. Traditional antivirus solutions have limitations in detecting mobile malware effectively. To address this, the paper proposes a deep learning-based approach using a convolutional neural network to detect and categorize malicious applications based on permission patterns. The solution achieves a 93% accuracy in identifying malware using a dataset of 2500 Android apps, including 2000 malicious and 500 benign ones.
The paper proposes an Android malware detection method using a convolutional neural network mixed-data model. The data includes API method calls and app permissions, represented using Word2vec technology for API calls and binary encoding for permissions. The input sequence is broken down into nibbles and normalized. The neural network has two parallel convolutional branches, one for each type of data, followed by fully connected layers. Each branch has two convolutional layers to map simple features used by the second layer to identify higher-level behavioral patterns. A pooling layer reduces data dimension, and the outputs from both branches are combined to determine the probability of an app being malware or benign.
Machine learning techniques are used to study and identify behavioral patterns in different malware families. Static analysis examines the content of malicious files without execution, while dynamic analysis analyzes their behavioral aspects by executing tasks like function call monitoring and information flow tracking. By leveraging machine learning, the static and dynamic artifacts of malware can predict the evolution of modern malware structures. This empowers systems to detect more complex malware attacks, which traditional methods struggle to predict effectively.
The usage of mobile devices has surged, with Android being the dominant operating system. Detecting malicious mobile applications among the millions available for download is crucial. Existing methods rely on static or dynamic features of APIs for Android malware detection. In this study, a novel approach is presented where Android malware and benign apks are converted into images based on their sizes. A convolutional neural network architecture is proposed for deep learning to identify Android malware. The framework achieved an impressive 97.76% accuracy in detecting malware using only classes.dex files of apks, requiring minimal information.
Exploring Capstone Project 1 Thesis at Sunway University (SU)? Count on us for support! Our thesis writing services in Malaysia cater to Malaysian students' needs. From navigating intricate topics like "CNN-Based Android Malware Detection" to crafting comprehensive theses, our assignment writer in Malaysia is here for you. Let our experts assist you in excelling in your coursework and conquering your Capstone Project 1 Thesis at SU, Malaysia.
