Description:
The use of smartphones and hand-held devices continues to increase with rapid development in underlying technology and widespread deployment of numerous applications including social network, email and financial transactions. Inevitably, malware attacks are shifting towards these devices. To detect mobile malware, features representing the characteristics of applications play a crucial role. In this work, we systematically studied the impact of all categories of features (i.e., permission, application programmers interface calls, inter component communication and dynamic features) of android applications in classifying a malware from benign applications. We identified the best combination of feature categories that yield better performance in terms of widely used metrics than blindly using all feature categories. We proposed a new technique to include contextual information in API calls into feature values and the study reveals that embedding such information enhances malware detection capability by a good margin. Information gain analysis shows that a significant number of features in ICC category is not relevant to malware prediction and hence, least effective. This study will be useful in designing better mobile malware detection system.
Description:
Due to its widespread use, with numerous applications deployed everyday, smartphones have become an inevitable target of the malware developers. This huge number of applications renders manual inspection of codes infeasible; as such, researchers have proposed several malware detection techniques based on automatic machine learning tools. Deep learning has gained a lot of attention from the malware researchers due to its ability of capture complex relationships among inputs and outputs. However, deep learning models depend largely on several hyper-parameters (i.e., learning rate, batch size, dropout rate). Hence, it is of utmost importance to analyze the effect of these parameters on classifier performance. In this paper, we systematically studied the effect of these parameters along with the effect of network architecture. We showed that building arbitrary deep networks does not always improve classifier performance. We also determined the combination of hyper-parameters that yields best result. This study will be useful in building better deep neural network based model for malware classification.
Description:
Industrial Internet of Things (IIoT) deploys edge devices to act as intermediaries between sensors and actuators and application servers or cloud services. Machine learning models have been widely used to thwart malware attacks in such edge devices. However, these models are vulnerable to adversarial attacks where attackers craft adversarial samples by introducing small perturbations to malware samples to fool a classifier to misclassify them as benign applications. Literature on deep learning networks proposes adversarial retraining as a defense mechanism where adversarial samples are combined with legitimate samples to retrain the classifier. However, existing works select such adversarial samples in a random fashion which degrades the classifier's performance. This work proposes two novel approaches for selecting adversarial samples to retrain a classifier. One, based on the distance from malware cluster center, and the other, based on a probability measure derived from a kernel based learning (KBL). Our experiments show that both of our sample selection methods outperform the random selection method and the KBL selection method improves detection accuracy by 6%. Also, while existing works focus on deep neural networks with respect to adversarial retraining, we additionally assess the impact of such adversarial samples on other classifiers and our proposed selective adversarial retraining approaches show similar performance improvement for these classifiers as well. The outcomes from the study can assist in designing robust security systems for IIoT applications.
Description:
Industrial Internet of Things (IIoT) deploys edge devices to act as intermediaries between sensors and actuators and application servers or cloud services. Machine learning models have been widely used to thwart malware attacks in such edge devices. However, these models are vulnerable to adversarial attacks where attackers craft adversarial samples by introducing small perturbations to malware samples to fool a classifier to misclassify them as benign applications. Literature on deep learning networks proposes adversarial retraining as a defense mechanism where adversarial samples are combined with legitimate samples to retrain the classifier. However, existing works select such adversarial samples in a random fashion which degrades the classifier's performance. This work proposes two novel approaches for selecting adversarial samples to retrain a classifier. One, based on the distance from malware cluster center, and the other, based on a probability measure derived from a kernel based learning (KBL). Our experiments show that both of our sample selection methods outperform the random selection method and the KBL selection method improves detection accuracy by 6%. Also, while existing works focus on deep neural networks with respect to adversarial retraining, we additionally assess the impact of such adversarial samples on other classifiers and our proposed selective adversarial retraining approaches show similar performance improvement for these classifiers as well. The outcomes from the study can assist in designing robust security systems for IIoT applications.
Description:
The increasing popularity and use of smartphones and hand-held devices have made them the most popular target for malware attackers. Researchers have proposed machine learning-based models to automatically detect malware attacks on these devices. Since these models learn application behaviors solely from the extracted features, choosing an appropriate and meaningful feature set is one of the most crucial steps for designing an effective mobile malware detection system. There are four categories of features for mobile applications. Previous works have taken arbitrary combinations of these categories to design models, resulting in sub-optimal performance. This thesis systematically investigates the individual impact of these feature categories on mobile malware detection systems. Feature categories that complement each other are investigated and categories that add redundancy to the feature space (thereby degrading the performance) are analyzed. In the process, the combination of feature categories that provides the best detection results is identified. Ensuring reliability and robustness of the above-mentioned malware detection systems is of utmost importance as newer techniques to break down such systems continue to surface. Adversarial attack is one such evasive attack that can bypass a detection system by carefully morphing a malicious sample even though the sample was originally correctly identified by the same system. Self-crafted adversarial samples can be used to retrain a model to defend against such attacks. However, randomly using too many such samples, as is currently done in the literature, can further degrade detection performance. This work proposed two intelligent approaches to retrain a classifier through the intelligent selection of adversarial samples. The first approach adopts a distance-based scheme where the samples are chosen based on their distance from malware and benign cluster centers while the second selects the samples based on a probability measure derived from a kernel-based learning method. The second method achieved a 6% improvement in terms of accuracy. To ensure practical deployment of malware detection systems, it is necessary to keep the real-world data characteristics in mind. For example, the benign applications deployed in the market greatly outnumber malware applications. However, most studies have assumed a balanced data distribution. Also, techniques to handle imbalanced data in other domains cannot be applied directly to mobile malware detection since they generate synthetic samples with broken functionality, making them invalid. In this regard, this thesis introduces a novel synthetic over-sampling technique that ensures valid sample generation. This technique is subsequently combined with a dynamic cost function in the learning scheme that automatically adjusts minority class weight during model training which counters the bias towards the majority class and stabilizes the model. This hybrid method provided a 9% improvement in terms of F1-score. Aiming to design a robust malware detection system, this thesis extensively studies machine learning-based mobile malware detection in terms of best feature category combination, resilience against evasive attacks, and practical deployment of detection models. Given the increasing technological advancements in mobile and hand-held devices, this study will be very useful for designing robust cybersecurity systems to ensure safe usage of these devices.
Description:
The increasing popularity and use of smartphones and hand-held devices have made them the most popular target for malware attackers. Researchers have proposed machine learning-based models to automatically detect malware attacks on these devices. Since these models learn application behaviors solely from the extracted features, choosing an appropriate and meaningful feature set is one of the most crucial steps for designing an effective mobile malware detection system. There are four categories of features for mobile applications. Previous works have taken arbitrary combinations of these categories to design models, resulting in sub-optimal performance. This thesis systematically investigates the individual impact of these feature categories on mobile malware detection systems. Feature categories that complement each other are investigated and categories that add redundancy to the feature space (thereby degrading the performance) are analyzed. In the process, the combination of feature categories that provides the best detection results is identified. Ensuring reliability and robustness of the above-mentioned malware detection systems is of utmost importance as newer techniques to break down such systems continue to surface. Adversarial attack is one such evasive attack that can bypass a detection system by carefully morphing a malicious sample even though the sample was originally correctly identified by the same system. Self-crafted adversarial samples can be used to retrain a model to defend against such attacks. However, randomly using too many such samples, as is currently done in the literature, can further degrade detection performance. This work proposed two intelligent approaches to retrain a classifier through the intelligent selection of adversarial samples. The first approach adopts a distance-based scheme where the samples are chosen based on their distance from malware and benign cluster centers while the second selects the samples based on a probability measure derived from a kernel-based learning method. The second method achieved a 6% improvement in terms of accuracy. To ensure practical deployment of malware detection systems, it is necessary to keep the real-world data characteristics in mind. For example, the benign applications deployed in the market greatly outnumber malware applications. However, most studies have assumed a balanced data distribution. Also, techniques to handle imbalanced data in other domains cannot be applied directly to mobile malware detection since they generate synthetic samples with broken functionality, making them invalid. In this regard, this thesis introduces a novel synthetic over-sampling technique that ensures valid sample generation. This technique is subsequently combined with a dynamic cost function in the learning scheme that automatically adjusts minority class weight during model training which counters the bias towards the majority class and stabilizes the model. This hybrid method provided a 9% improvement in terms of F1-score. Aiming to design a robust malware detection system, this thesis extensively studies machine learning-based mobile malware detection in terms of best feature category combination, resilience against evasive attacks, and practical deployment of detection models. Given the increasing technological advancements in mobile and hand-held devices, this study will be very useful for designing robust cybersecurity systems to ensure safe usage of these devices.