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:
Obfuscated Memory Malware (OMM) presents significant threats to interconnected systems, including smart city applications, for its ability to evade detection through concealment tactics. Existing OMM detection methods primarily focus on binary detection. Their multiclass versions consider a few families only and, thereby, fail to detect much existing and emerging malware. Moreover, their large memory size makes them unsuitable to be executed in resource-constrained embedded/IoT devices. To address this problem, in this paper, we propose a multiclass but lightweight malware detection method capable of identifying recent malware and is suitable to execute in embedded devices. For this, the method considers a hybrid model by combining the feature-learning capabilities of convolutional neural networks with the temporal modeling advantage of bidirectional long short-term memory. The proposed architecture exhibits compact size and fast processing speed, making it suitable for deployment in IoT devices that constitute the major components of smart city systems. Extensive experiments with the recent CIC-Malmem-2022 OMM dataset demonstrate that our method outperforms other machine learning-based models proposed in the literature in both detecting OMM and identifying specific attack types. Our proposed method thus offers a robust yet compact model executable in IoT devices for defending against obfuscated malware.
Description:
Obfuscated Memory Malware (OMM) presents significant threats to interconnected systems, including smart city applications, for its ability to evade detection through concealment tactics. Existing OMM detection methods primarily focus on binary detection. Their multiclass versions consider a few families only and, thereby, fail to detect much existing and emerging malware. Moreover, their large memory size makes them unsuitable to be executed in resource-constrained embedded/IoT devices. To address this problem, in this paper, we propose a multiclass but lightweight malware detection method capable of identifying recent malware and is suitable to execute in embedded devices. For this, the method considers a hybrid model by combining the feature-learning capabilities of convolutional neural networks with the temporal modeling advantage of bidirectional long short-term memory. The proposed architecture exhibits compact size and fast processing speed, making it suitable for deployment in IoT devices that constitute the major components of smart city systems. Extensive experiments with the recent CIC-Malmem-2022 OMM dataset demonstrate that our method outperforms other machine learning-based models proposed in the literature in both detecting OMM and identifying specific attack types. Our proposed method thus offers a robust yet compact model executable in IoT devices for defending against obfuscated malware.
Description:
The revolution in online criminal activities and malicious software (malware) has posed a serious challenge in malware forensics. Malicious attacks have become more organized and purposefully directed. With cybercrimes escalating to great heights in quantity as well as in sophistication and stealth, the main challenge is to detect hidden and obfuscated malware. Malware authors use a variety of obfuscation methods and specialized stealth techniques of information hiding to embed malicious code, to infect systems and to thwart any attempt to detect them, specifically with the use of commercially available anti-malware engines. This has led to the situation of zero-day attacks, where malware inflict systems even with existing security measures. The aim of this thesis is to address this situation by proposing a variety of novel digital forensic and data mining techniques to automatically detect hidden and obfuscated malware. Anti-malware engines use signature matching to detect malware where signatures are generated by human experts by disassembling the file and selecting pieces of unique code. Such signature based detection works effectively with known malware but performs poorly with hidden or unknown malware. Code obfuscation techniques, such as packers, polymorphism and metamorphism, are able to fool current detection techniques by modifying the parent code to produce offspring copies resulting in malware that has the same functionality, but with a different structure. These evasion techniques exploit the drawbacks of traditional malware detection methods, which take current malware structure and create a signature for detecting this malware in the future. However, obfuscation techniques aim to reduce vulnerability to any kind of static analysis to the determent of any reverse engineering process. Furthermore, malware can be hidden in file system slack space, inherent in NTFS file system based partitions, resulting in malware detection that even more difficult.
Description:
The revolution in online criminal activities and malicious software (malware) has posed a serious challenge in malware forensics. Malicious attacks have become more organized and purposefully directed. With cybercrimes escalating to great heights in quantity as well as in sophistication and stealth, the main challenge is to detect hidden and obfuscated malware. Malware authors use a variety of obfuscation methods and specialized stealth techniques of information hiding to embed malicious code, to infect systems and to thwart any attempt to detect them, specifically with the use of commercially available anti-malware engines. This has led to the situation of zero-day attacks, where malware inflict systems even with existing security measures. The aim of this thesis is to address this situation by proposing a variety of novel digital forensic and data mining techniques to automatically detect hidden and obfuscated malware. Anti-malware engines use signature matching to detect malware where signatures are generated by human experts by disassembling the file and selecting pieces of unique code. Such signature based detection works effectively with known malware but performs poorly with hidden or unknown malware. Code obfuscation techniques, such as packers, polymorphism and metamorphism, are able to fool current detection techniques by modifying the parent code to produce offspring copies resulting in malware that has the same functionality, but with a different structure. These evasion techniques exploit the drawbacks of traditional malware detection methods, which take current malware structure and create a signature for detecting this malware in the future. However, obfuscation techniques aim to reduce vulnerability to any kind of static analysis to the determent of any reverse engineering process. Furthermore, malware can be hidden in file system slack space, inherent in NTFS file system based partitions, resulting in malware detection that even more difficult.
Description:
The Internet is a decentralized structure that offers speedy communication, has a global reach and provides anonymity, a characteristic invaluable for committing illegal activities. In parallel with the spread of the Internet, cybercrime has rapidly evolved from a relatively low volume crime to a common high volume crime. A typical example of such a crime is the spreading of spam emails, where the content of the email tries to entice the recipient to click a URL linking to a malicious Web site or downloading a malicious attachment. Analysts attempting to provide intelligence on spam activities quickly find that the volume of spam circulating daily is overwhelming; therefore, any intelligence gathered is representative of only a small sample, not of the global picture. While past studies have looked at automating some of these analyses using topic-based models, i.e. separating email clusters into groups with similar topics, our preliminary research investigates the usefulness of applying authorship-based models for this purpose. In the first phase, we clustered a set of spam emails using an authorship-based clustering algorithm. In the second phase, we analysed those clusters using a set of linguistic, structural and syntactic features. These analyses reveal that emails within each cluster were likely written by the same author, but that it is unlikely we have managed to group together all spam produced by each group. This problem of high purity with low recall, has been faced in past authorship research. While it is also a limitation of our research, the clusters themselves are still useful for the purposes of automating analysis, because they reduce the work needing to be performed. Our second phase revealed useful information on the group that can be utilized in future research for further analysis of such groups, for example, identifying further linkages behind spam campaigns.
A critical review of intrusion detection systems in the internet of things : techniques, deployment strategy, validation strategy, attacks, public datasets and challenges
Description:
Banking Malware, has become a popular and ever more prevalent mechanism to monetise malware development. Since the development of the Zeus malware kit in 2007, the frequency and complexity of banking malware has been increasing. Developing a good understanding of the operation of a malware family is a first step in the reverse engineering required to create tools to extract the malware configuration, which is used in the remediation of malware infrastructure. This reverse engineering process in recent years has become increasingly challenging. This manuscript provides a brief summary of the reverse engineering of banking malware families over a two year period and emphasises the anti-analysis techniques employed by the authors of six families of banking malware. The manuscript presents this analysis, and examines trends in the development of these anti-analysis techniques.
Description:
In this paper, we propose a five-step approach to detect obfuscated malware by investigating the structural and behavioural features of API calls. We have developed a fully automated system to disassemble and extract API call features effectively from executables. Using n-gram statistical analysis of binary content, we are able to classify if an executable file is malicious or benign. Our experimental results with a dataset of 242 malwares and 72 benign files have shown a promising accuracy of 96.5% for the unigram model. We also provide a preliminary analysis by our approach using support vector machine (SVM) and by varying n-values from 1 to 5, we have analysed the performance that include accuracy, false positives and false negatives. By applying SVM, we propose to train the classifier and derive an optimum n-gram model for detecting both known and unknown malware efficiently.
Description:
Due to increase in intrusion activities over internet, many intrusion detection systems are proposed to detect abnormal activities, but most of these detection systems suffer a common problem which is producing a high number of alerts and a huge number of false positives. As a result, normal activities could be classified as intrusion activities. This paper examines different data mining techniques that could minimize both the number of false negatives and false positives. C5 classifier’s effectiveness is examined and compared with other classifiers. Results should that false negatives are reduced and intrusion detection has been improved significantly. A consequence of minimizing the false positives has resulted in reduction in the amount of the false alerts as well. In this study, multiple classifiers have been compared with C5 decision tree classifier using NSL_KDD dataset and results have shown that C5 has achieved high accuracy and low false alarms as an intrusion detection system.
Description:
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)