Speaker
Description
With the increasing number of cyber threats, malware detection has become a critical challenge in cybersecurity. Traditional detection techniques, based on behavioral analysis and signature creation, present significant challenges due to their time-consuming nature and limited effectiveness against new, unknown threats. This paper explores the optimization of artificial intelligence (AI) for malware detection, leveraging acceleration on mobile neural processing units (NPUs). The primary goal of this study is to develop and evaluate methods that enhance machine learning models in terms of computational efficiency and detection accuracy in resource-constrained environments.
The research methodology involved designing the initial model on a high-performance computing system to determine the optimal network size that balances processing speed and detection precision across various devices. A more precise version of the model was subsequently trained on more powerful hardware, followed by the application of optimization techniques to adapt it for execution on mobile NPUs.
The study examines the performance of different NPU architectures and assesses the impact of operating systems on the efficiency and speed of AI-based detection models. Preliminary findings indicate that the applied optimizations have led to satisfactory results in detecting malicious code within binary files on Windows systems.
This research contributes to the development of AI deployment methods on mobile devices, which can significantly enhance end-user security without relying on external cloud-based solutions. Future studies aim to expand the analysis to software running on other platforms.
