A deep learning system for health care IoT and smartphone malware detection

Muhammad Amin, Duri Shehwar, Abrar Ullah, Teresa Guarda, Tamleek Ali Tanveer*, Sajid Anwar

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

19 Citations (Scopus)


The use of smart and connected devices, such as Android and Internet of Things (IoT) have increased exponentially. In the last 10 years, mobiles and IoT devices have surpassed PC’s utilization. Android hosts an array of connected sensors like IoT. It has transformed a simple gadget into a hub of mobile phone with IoT. With a high number of clients and enormous assortment of Android applications it has been an appealing target for many security threats including malware attacks. To monitor a host of the applications that runs on Android and IoT devices, this study employs a deep learning based feature detector for malware detection which can easily be trained and be used with different classifiers to assess an application’s behavior. The features learnt by the detector can be reused to transfer their learning to any future endeavors toward malware detection. To test the accuracy and effectiveness of the feature detector we test it in two phases: (i) first the features extracted are fed to a fully connected network (FCN) with Softmax activation and in (ii) second scheme we use recurrent layers of attentions to classify the Applications either as malicious or benign. Our findings reveal that the proposed feature detector achieves significant results with an F1-Score of 98.97% and an accuracy of 98%.

Original languageEnglish
Pages (from-to)11283–11294
Number of pages12
JournalNeural Computing and Applications
Early online date6 Nov 2020
Publication statusPublished - Jul 2022


  • Deep learning
  • Feature detector
  • Malware detection

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence


Dive into the research topics of 'A deep learning system for health care IoT and smartphone malware detection'. Together they form a unique fingerprint.

Cite this