The availability of big data and affordable hardware have enabled the applications of deep learning on different tasks. With respect to security, several attempts have been made to transfer deep learning’s application from the domain of image recognition or natural language processing into malware detection. In this study, we propose AdMat - a simple yet effective framework to characterize Android applications by treating them as images. The novelty of our study lies in the construction of an adjacency matrix for each application. These matrices act as “input images” to the Convolutional Neural Network model, allowing it to learn to differentiate benign and malicious apps, as well as malware families. During the experiment, we found that AdMat was able to adapt to a variety of training ratios and achieve the average detection rate of 98.26% in different malware datasets. In classification tasks, it also successfully recognized over 97.00% of different malware families with limited number of training data.
29 October 2017 in Security and Communication Networks
We present an arms race between rooting detection and rooting evasion. We investigate different methods to detect rooted device at both Java and native level and evaluate the counterattack from major hooking tools. To this end, an extensive study of Android rooting has been conducted, which includes the techniques to root the device and make it invisible to the detection of mobile antimalware product. We then analyze the evasion loopholes and in turn enhance our rooting detection tool. We also apply evasion techniques on rooted device and compare our work with 92 popular root checking applications and 18 banking and finance applications. Results show that most of them do not suffice and can be evaded through API hooking or static file renaming. Furthermore, over 28000 Android applications have been analyzed and evaluated in order to diagnose the characteristics of rooting in recent years. Our study shows that rooting has become more and more prevalent as an inevitable trend, and it raises big security concerns regarding detection and evasion. As a proof of concept, we have published our rooting detection application to Google Play Store to demonstrate the work presented in this paper.
Critical infrastructures have recently been integrated with digital controls to support intelligent decision making. Although this integration provides various benefits and improvements, it also exposes the system to new cyberattacks. In particular, the injection of false data and commands into communication is one of the most common and fatal cyberattacks in critical infrastructures. Hence, in this paper, we investigate the effectiveness of machine-learning algorithms in detecting False Data Injection Attacks (FDIAs). In particular, we focus on two of the most widely used critical infrastructures, namely power systems and water treatment plants. This study focuses on tackling two key technical issues: (1) finding the set of best features under a different combination of techniques and (2) resolving the class imbalance problem using oversampling methods. We evaluate the performance of each algorithm in terms of time complexity and detection accuracy to meet the time-critical requirements of critical infrastructures. Moreover, we address the inherent skewed distribution problem and the data imbalance problem commonly found in many critical infrastructure datasets. Our results show that the considered minority oversampling techniques can improve the Area Under Curve (AUC) of GradientBoosting, AdaBoost, and kNN by 10–12%.
Container technology plays an essential role in many Information and Communications Technology (ICT) systems. However, containers face a diversity of threats caused by vulnerable packages within container images. Previous vulnerability scanning solutions for container images are inadequate. These solutions entirely depend on the information extracted from package managers. As a result, packages installed directly from the source code compilation, or packages downloaded from the repository, etc., are ignored. We introduce DAVS–A Dockerfile analysis-based vulnerability scanning framework for OCI-based container images to deal with the limitations of existing solutions. DAVS performs static analysis using file extraction based on Dockerfile information to obtain the list of Potentially Vulnerable Files (PVFs). The PVFs are then scanned to figure out the vulnerabilities in the target container image. The experimental shows the outperform of DAVS on detecting Common Vulnerabilities and Exposures (CVE) of 10 known vulnerable images compared to Clair– the most popular container image scanning project. Moreover, DAVS found that 68% of real-world container images are vulnerable from different image registries.
1 August 2015 in International Journal of Distributed Sensor Networks
The dramatically increasing number of connected things based on Internet protocol is leading to a new concept called the Internet of Things (IoT). The Internet-integrated wireless sensor network has recently become one of the most important service targets in IoT field. To provide secure IoT services, the IETF proposed using Datagram Transport Layer Security (DTLS) as a de facto security protocol. In this paper, we examine problems that can happen when applying the DTLS protocol to the IoT, which comprises constrained devices and constrained networks. To solve the problems at hand, we separate the DTLS protocol into a handshake phase (i.e., establishment phase) and an encryption phase (i.e., transmission phase). Our approach enhances the performance of both device and network by using a way to delegate the DTLS handshake phase. The proposed scheme supports secure end-to-end communication despite using delegation.
Deepfake content-including audio, video, images, and text-synthesized or modified using artificial intelligence is designed to convincingly mimic real content. As deepfake generation technology advances, detecting deepfake content presents significant challenges. While recent progress has been made in detection techniques, identifying deepfake audio remains particularly challenging. Previous approaches have attempted to capture deepfake features by combining video and audio content; however, these methods are ineffective when video and audio are mismatched due to occlusion. To address this, we proposes a novel dual-channel deepfake audio detection model that leverages the direct and reverberant components extracted from raw audio signals, focusing exclusively on audio-based detection without reliance on video content. Across various datasets, including ASVspoof2019, FakeAVCeleb, and sport press conference datasets collected by our group, the proposed dual-channel model demonstrates significant improvements in quantitative metrics such as equal error rate and area under the curve. The implementation is available at https://github.com/gunwoo5034/Dual-Channel-Audio-Deepfake-Detection.