Efficient IoT Malware Detection Using Convolution Neural Network and View-Invariant Block

The numerous malware concerns in IoT devices, including manipulation through open-source platforms, require swift identification due to their potential for intrusion and data ex-filtration. Presently, the detection of sophisticated malware attacks poses a substantial challenge to the research and development community due to the evolving tactics employed by hackers over time. To address this evolving landscape, a novel malware detection framework is proposed. In its primary stage, we present an inventive solution that entails the representation of original malware binary files as color images. The employment of the view-invariant model yields a diverse array of feature maps. This entails generating a sequence of images from multiple views, necessitating the application of multi-dimension spatial rotations to these images. To encompass a wider spectrum of sequential data, a repertoire of twelve distinct rotational matrices is employed. Subsequent to this phase, a technique known as local-global feature aggregation is harnessed to amalgamate local intricacies with global contextual information, thereby enriching the comprehension of visual data. Rigorous experimentation and meticulous ablation studies substantiate the efficacy and superiority of the proposed model.

I. Introduction

The swift expansion of malicious software, intentionally crafted to covertly undermine computer systems, has experienced noteworthy acceleration. A diverse array of malicious software, including worms, trojans, botnets, ransom ware, and spyware, among others, is employed for such nefarious purposes. Malware invariably infiltrates systems or devices sneakily, bypassing the authorization and intention of their rightful owners. The proliferation of the internet has facilitated the widespread adoption of Internet of Things (IoT) devices, presenting a unique set of security challenges [1]. The inter-connectivity inherent in IoT devices has magnified these challenges, with the vast quantities of data generated by these devices attracting the attention of malicious actors. Various forms of malware are employed to gain access to these systems, enabling the collection of vital information [2]. Consequently, the detection of malware has emerged as a crucial concern, assuming paramount importance in safeguarding the security of IoT devices [3].