Deep Neural Networks for Automatic Android Malware Detection

Hou, Shifu; Saas, Aaron; Chen, Lingwei; Ye, Yanfang; Bourlai, Thirimachos

doi:10.1145/3110025.3116211

Cited by 39 publications

(24 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…( Yuan et al, 2016 ) in all other metrics, while utilizing more samples for the experiments. DL-Droid also outperforms Maldozer ( Karbab et al, 2017 ), Deep4MalDroid ( Hou et al, 2016 ), AutoDroid ( Hou et al, 2017 ) and the CNN approach presented in ( McLaughlin et al, 2017 ). It is interesting to note that, just like in Deep4MalDroid and Auto-Droid, the number of the optimum hidden layers for DL-Droid is three.…”

Section: Comparison Of the Performance Of The Deep Learning Classifiementioning

confidence: 89%

DL-Droid: Deep learning based android malware detection using real devices

Alzaylaee

Yerima

Sezer

2020

Computers & Security

316

141

View full text Add to dashboard Cite

a b s t r a c tThe Android operating system has been the most popular for smartphones and tablets since 2012. This popularity has led to a rapid raise of Android malware in recent years. The sophistication of Android malware obfuscation and detection avoidance methods have significantly improved, making many traditional malware detection methods obsolete. In this paper, we propose DL-Droid, a deep learning system to detect malicious Android applications through dynamic analysis using stateful input generation. Experiments performed with over 30,0 0 0 applications (benign and malware) on real devices are presented. Furthermore, experiments were also conducted to compare the detection performance and code coverage of the stateful input generation method with the commonly used stateless approach using the deep learning system. Our study reveals that DL-Droid can achieve up to 97.8% detection rate (with dynamic features only) and 99.6% detection rate (with dynamic + static features) respectively which outperforms traditional machine learning techniques. Furthermore, the results highlight the significance of enhanced input generation for dynamic analysis as DL-Droid with the state-based input generation is shown to outperform the existing state-of-the-art approaches.

show abstract

Section: Comparison Of the Performance Of The Deep Learning Classifiementioning

confidence: 89%

DL-Droid: Deep learning based android malware detection using real devices

Alzaylaee

Yerima

Sezer

2020

Computers & Security

316

141

View full text Add to dashboard Cite

show abstract

“…Studies [3], [33], [46], [194] all considered restricted API calls and suspicious API calls as features to detect malapps. Instead of using API calls directly, Hou et al [82] further categorized the API calls belonging to the same method in the smali code into a block, namely API call block. Their experimental results showed that the API call block outperformed using API calls directly in Android malapp detection.…”

Section: ) Apimentioning

confidence: 99%

“…They tested the method on a recent dataset composed of 7100 real-world mobile apps, obtaining an accuracy ranging between 0.85 and 0.95. Based on the extracted system calls, Hou et al [82] constructed the weighted directed graphs and then applied a deep learning framework for newly unknown Android malapp detection. They evaluated the performance of their proposed Deep4MalDroid.…”

Section: ) System Callmentioning

confidence: 99%

Constructing Features for Detecting Android Malicious Applications: Issues, Taxonomy and Directions

et al. 2019

View full text Add to dashboard Cite

The number of applications (apps) available for smart devices or Android based IoT (Internet of Things) has surged dramatically over the past few years. Meanwhile, the volume of ill-designed or malicious apps (malapps) has been growing explosively. To ensure the quality and security of the apps in the markets, many approaches have been proposed in recent years to discriminate malapps from benign ones. Machine learning is usually utilized in classification process. Accurately characterizing apps' behaviors, or so-called features, directly affects the detection results with machine learning algorithms. Android apps evolve very fast. The size of current apps has become increasingly large and the behaviors of apps have become increasingly complicated. The extracting effective and representative features from apps is thus an ongoing challenge. Although many types of features have been extracted in existing work, to the best of our knowledge, no work has systematically surveyed the features constructed for detecting Android malapps. In this paper, we are motivated to provide a clear and comprehensive survey of the state-of-the-art work that detects malapps by characterizing behaviors of apps with various types of features. Through the designed criteria, we collect a total of 1947 papers in which 236 papers are used for the survey with four dimensions: the features extracted, the feature selection methods employed if any, the detection methods used, and the scale of evaluation performed. Based on our in-depth survey, we highlight the issues of exploring effective features from apps, provide the taxonomy of these features and indicate the future directions.

show abstract

“…The proposed model outperforms various traditional ML models and achieves F1 score of 95.05%. In [350], a novel feature representation and two DL based malware detectors for android systems is proposed. The detection framework uses block of API calls as features instead of simple API calls which makes it harder for the attacker to evade detection.…”

Section: ) Deep Neural Network (Dnn)mentioning

confidence: 99%