.png){kind=icon}
.png){kind=icon}
Return
HFE-DDL: hybrid feature extraction and dual deep learning for android code vulnerability detection
.png){kind=icon}
PRE
AI
A
H
M
T
N
R
DOI：10.1007/s10586-026-06460-8.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The growing complexity of Android apps demands robust vulnerability detection to mitigate security risks. However, existing methods struggle to integrate lexical patterns with semantic structures, often suffering from overfitting and poor scalability. This paper introduces HFE-DDL, an Android code vulnerability detection approach that leverages hybrid feature extraction with a dual deep learning model. The proposed methodology integrates statistical feature extraction based on optimized TF-IDF vectorization and semantic modeling via sequence token embeddings using a regularized dual deep learning network to recognize vulnerable-contributing code patterns. This dual-level representation enables a more in-depth analysis of code vulnerabilities. This framework advances the field of automated vulnerability detection by integrating dual deep learning and hybrid code representations, offering practical implications for secure software engineering. The performance of the proposed approach was evaluated on a large-scale dataset comprising over 15 million labeled code samples from more than 15,000 Android apps. Experimental results demonstrate the superiority of HFE-DDL over the state-of-the-art method in vulnerability detection tasks. The results show that HFE-DDL has potential as a tool for proactively identifying vulnerabilities during development, aiming to produce more secure Android applications. The source code can be accessed via GitHub: https://github.com/Hakim-Abdo/HFE-DDL .
Keywords:
Code vulnerability
Android vulnerability detection
Source code representations
Deep learning
Journal
C
IF:
4.1
Papers:
4.8K
Citations:
7.5K
Organization
Cited Papers