Hamid Bostani

Adversarial Training (AT) has been widely applied to harden learning-based classifiers against adversarial evasive attacks. However, its effectiveness in identifying and strengthening vulnerable areas of the model's decision space while maintaining high performance on clean data of malware classifiers remains an under-explored area. In this context, the robustness that AT achieves has often been assessed against unrealistic or weak adversarial attacks, which negatively affect performance on clean data and are arguably no longer threats. Previous work seems to suggest robustness is a task-dependent property of AT. We instead argue it is a more complex problem that requires exploring AT and the intertwined roles played by certain factors within data, feature representations, classifiers, and robust optimization settings, as well as proper evaluation factors, such as the realism of evasion attacks, to gain a true sense of AT's effectiveness. In our paper, we address this gap by systematically exploring the role such factors have in hardening malware classifiers through AT. Contrary to recent prior work, a key observation of our research and extensive experiments confirm the hypotheses that all such factors influence the actual effectiveness of AT, as demonstrated by the varying degrees of success from our empirical analysis. We identify five evaluation pitfalls that affect state-of-the-art studies and summarize our insights in ten takeaways to draw promising research directions toward better understanding the factors' settings under which adversarial training works at best.

In the battle against evolving cyber threats, robust malware detection with machine learning requires more than just advanced algorithms—it demands high-quality data. This article emphasizes data quality's role and how leveraging coresets—constructed from key samples that preserve the datasets’ properties—to enrich training sets can enhance malware classifiers’ resilience.

Machine Learning (ML) promises to enhance the efficacy of Android Malware Detection (AMD); however, ML models are vulnerable to realistic evasion attacks—crafting realizable Adversarial Examples (AEs) that satisfy Android malware domain constraints. To eliminate ML vulnerabilities, defenders aim to identify susceptible regions in the feature space where ML models are prone to deception. The primary approach to identifying vulnerable regions involves investigating realizable AEs, but generating these feasible apps poses a challenge. For instance, previous work has relied on generating either feature-space norm-bounded AEs or problem-space realizable AEs in adversarial hardening. The former is efficient but lacks full coverage of vulnerable regions while the latter can uncover these regions by satisfying domain constraints but is known to be time-consuming. To address these limitations, we propose an approach to facilitate the identification of vulnerable regions. Specifically, we introduce a new interpretation of Android domain constraints in the feature space, followed by a novel technique that learns them. Our empirical evaluations across various evasion attacks indicate effective detection of AEs using learned domain constraints, with an average of 89.6%. Furthermore, extensive experiments on different Android malware detectors demonstrate that utilizing our learned domain constraints in Adversarial Training (AT) outperforms other AT-based defenses that rely on norm-bounded AEs or state-of-the-art non-uniform perturbations. Finally, we show that retraining a malware detector with a wide variety of feature-space realizable AEs results in a 77.9% robustness improvement against realizable AEs generated by unknown problem-space transformations, with up to 70x faster training than using problem-space realizable AEs.

Machine learning (ML) has demonstrated significant advancements in Android malware detection (AMD); however, the resilience of ML against realistic evasion attacks remains a major obstacle for AMD. One of the primary factors contributing to this challenge is the scarcity of reliable generalizations. Malware classifiers with limited generalizability tend to overfit spurious correlations derived from biased features. Consequently, adversarial examples (AEs), generated by evasion attacks, can modify these features to evade detection. In this study, we propose a domain adaptation technique to improve the generalizability of AMD by aligning the distribution of malware samples and AEs. Specifically, we utilize meaningful feature dependencies, reflecting domain constraints in the feature space, to establish a robust feature space. Training on the proposed robust feature space enables malware classifiers to learn from predefined patterns associated with app functionality rather than from individual features. This approach helps mitigate spurious correlations inherent in the initial feature space. Our experiments conducted on DREBIN, a renowned Android malware detector, demonstrate that our approach surpasses the state-of-the-art defense, Sec-SVM, when facing realistic evasion attacks. In particular, our defense can improve adversarial robustness by up to 55% against realistic evasion attacks compared to Sec-SVM.

Over the last decade, researchers have extensively explored the vulnerabilities of Android malware detectors to adversarial examples through the development of evasion attacks; however, the practicality of these attacks in real-world scenarios remains arguable. The majority of studies have assumed attackers know the details of the target classifiers used for malware detection, while in reality, malicious actors have limited access to the target classifiers. This paper introduces EvadeDroid, a problem-space adversarial attack designed to effectively evade black-box Android malware detectors in real-world scenarios. EvadeDroid constructs a collection of problem-space transformations derived from benign donors that share opcode-level similarity with malware apps by leveraging an n-gram-based approach. These transformations are then used to morph malware instances into benign ones via an iterative and incremental manipulation strategy. The proposed manipulation technique is a query-efficient optimization algorithm that can find and inject optimal sequences of transformations into malware apps. Our empirical evaluations, carried out on 1K malware apps, demonstrate the effectiveness of our approach in generating real-world adversarial examples in both soft- and hard-label settings. Our findings reveal that EvadeDroid can effectively deceive diverse malware detectors that utilize different features with various feature types. Specifically, EvadeDroid achieves evasion rates of 80%–95% against DREBIN, Sec-SVM, ADE-MA, MaMaDroid, and Opcode-SVM with only 1–9 queries. Furthermore, we show that the proposed problem-space adversarial attack is able to preserve its stealthiness against five popular commercial antiviruses with an average of 79% evasion rate, thus demonstrating its feasibility in the real world.

On the modern web, trackers and advertisers frequently construct and monetize users’ detailed behavioral profiles without consent. Despite various studies on web tracking mechanisms and advertisements, there has been no rigorous study focusing on websites targeted at children. To address this gap, we present a measurement of tracking and (targeted) advertising on websites directed at children. Motivated by the lack of a comprehensive list of child-directed (i.e., targeted at children) websites, we first build a multilingual classifier based on web page titles and descriptions. Applying this classifier to over two million pages from the Common Crawl dataset, we compile a list of two thousand child-directed websites. Crawling these sites from five vantage points, we measure the prevalence of trackers, fingerprinting scripts, and advertisements. Our crawler detects ads displayed on child-directed websites and determines if ad targeting is enabled by scraping ad disclosure pages whenever available. Our results show that around 90% of child-directed websites embed one or more trackers, and about 27% contain targeted advertisements—a practice that should require verifiable parental consent. Next, we identify improper ads on child-directed websites by developing an ML pipeline that processes both images and text extracted from ads. The pipeline allows us to run semantic similarity queries for arbitrary search terms, revealing ads that promote services related to dating, weight loss, and mental health, as well as ads for sex toys and flirting chat services. Some of these ads feature repulsive, sexually-explicit and highly-inappropriate imagery. In summary, our findings indicate a trend of non-compliance with privacy regulations and troubling ad safety practices among many advertisers and child-directed websites. To ensure the protection of children and create a safer online environment, regulators and stakeholders must adopt and enforce more stringent measures.

In this chapter, in order to show the efficiency of OPF in the intrusion detection systems (IDSs) and also in the large-scale problems, we introduce five hybrid and modified OPFs as follows: (a) a modified OPF using unsupervised learning and social network concept; (b) a hybrid IDS using unsupervised OPF based on MapReduce approach; (c) a hybrid IDS using a modified OPF (MOPF) and selected input features; (d) a modified OPF using Markov cluster process algorithm; and (e) a modified OPF based on the coreset concept. Furthermore, the MOPF-based IDS is improved in the last section as a contribution by using an outperformed clustering algorithm.

Optimum-path forest (OPF) is one of the efficient graph-based frameworks that can determine the patterns of input dataset by extracting the optimal partitions of graph obtained through encoding data into a graph. Since OPF was introduced based on simple assumptions without considering the requirements of large-scale problems, this machine learning is an effective algorithm only for a reasonable size of input datasets. To provide a scalable OPF, this study introduces a strong coreset for accelerating OPF algorithm. Applying this approach can expedite OPF procedure, especially when it is working on massive datasets. Accordingly, a novel algebra is developed to represent the problem of OPF as an optimization problem for the proposed coreset definition. A novel coreset construction algorithm that can approximate the OPF solutions is subsequently proposed in order to improve the OPF construction speed. The simulation results of diverse experiments on various benchmark datasets illustrate computation gain and superiority of the proposed algorithm in terms of the construction and classification speeds as compared to the original algorithm while displaying reliably accurate performance. The presented coreset construction algorithm performs the training and testing phases of OPF up to 6.1 and 4.9 times faster than before, respectively.

Optimum-path forest (OPF) is an effective graph-based machine learning that simplifies the pattern recognition problems into the partitioning the corresponding derived graphs of the input datasets. The amounts of the samples in the input datasets and, consequently the size of the node set of their corresponding derived graphs has a major effect on the speed of OPF. In this study a novel version of OPF is introduced which utilizes coreset approach for reducing the scale of the input dataset. From the aspect of the computational geometry, coreset is a small set of points that includes the best representative points of the original point set with regard to a geometric objective function. Our method finds the most informative vertices (samples) by proposing a novel incremental coreset construction algorithm. The experimental results of the proposed method reduces the input data samples, and the execution times of the construction and the classification phases of OPF by 80%, 60%, and 12%, respectively, in contrast to the traditional OPF.

Internet of Things (IoT) is a novel paradigm in computer networks in which resource-constrained objects connect to unreliable Internet by using a wide range of technologies. The insecure nature of the Internet and wireless sensor networks, that are the main components of IoT, make IoT vulnerable to different attacks, especially routing attacks (as insider attacks). A novel real-time hybrid intrusion detection framework is proposed in this study that consists of anomaly-based and specification-based intrusion detection modules for detecting two well-known routing attacks in IoT called sinkhole and selective-forwarding attacks. For this purpose, the specification-based intrusion detection agents, that are located in the router nodes, analyze the behavior of their host nodes and send their local results to the root node through normal data packets. In addition, an anomaly-based intrusion detection agent, that is located in the root node, employs the unsupervised optimum-path forest algorithm for projecting clustering models by using incoming data packets. This agent, which is based on the MapReduce architecture, can work in a distributed platform for projecting clustering models and consequently parallel detecting of anomalies as a global detection approach. The proposed method makes decision about suspicious behavior by using a voting mechanism. Notably, the proposed method is also extended to detect wormhole attack. The deployment of the hybrid proposed model is investigated in a smart-city scenario by an existing platform, as well. The free network's scale and the ability to identify malicious nodes are two key features of the proposed framework that are evaluated through different experiments in this study. The experimental results of simulated scenarios showed that the proposed hybrid method can achieve true positive rate of 76.19% and false positive rate of 5.92% when both sinkhole and selective-forwarding attacks were launched simultaneously. These rates in detecting wormhole attack are 96.02% and 2.08%, respectively.

Optimum-path forest (OPF) is a graph-based machine learning method that can overcome some limitations of the traditional machine learning algorithms that have been used in intrusion detection systems. This paper presents a novel approach for intrusion detection using a modified OPF (MOPF) algorithm for improving the performance of traditional OPF in terms of detection rate (DR), false alarm rate (FAR), and time of execution. To address the problem of scalability in large datasets and also for achieving high attack recognition rates, the proposed framework employs the k-means clustering algorithm, as a partitioning module, for generating different homogeneous training subsets from original heterogeneous training samples. In the proposed MOPF algorithm, the distance between unlabeled samples and the root (prototype) of every sample in OPF is also considered in classifying unlabeled samples with the aim of improving the accuracy rate of traditional OPF algorithm. Moreover, the centrality and the prestige concepts in the social network analysis are employed in a pruning module for determining the most informative samples in training subsets to speed up the traditional OPF algorithm. The experimental results on NSL-KDD dataset show that the proposed method performs better than traditional OPF in terms of accuracy rate, DR, FAR, and cost per example (CPE) evaluation metrics.

Internet of things (IoT) is a novel emerging approach in computer networks wherein all heterogeneous objects around us, which usually are resource-constrained objects, can connect to each other and also the Internet by using a broad range of technologies. IoT is a hybrid network which includes the Internet and also wireless sensor networks (WSNs) as the main components of IoT; so, implementing security mechanisms in IoT seems necessary. This paper introduces a novel intrusion detection architecture model for IoT that provides the possibility of distributed detection. The proposed hybrid model uses anomaly and misuse intrusion detection agents based on the supervised and unsupervised optimum-path forest models for providing the ability to detect internal and externals attacks, simultaneously. The number of input features to the proposed classifier is reduced by a hybrid feature selection algorithm, as well. The experimental results of simulated scenarios show the superior performance of proposed security mechanism in multi-faceted detection.

Today, detecting anomalous traffic and preventing it in computer networks has become increasingly important for the community of security researchers. An intrusion detection system (IDS) is an effective tool for reaching high security. This is a software tool for analyzing system behavior or network traffic as input data to detect deviations from normal behavior. With the development of computer networks, highdimensional input data analysis has become a huge problem in IDSs. One solution for overcoming this problem is feature selection, which is a process for selecting an optimal subset of features. Populationbased heuristic search algorithms have been widely used for this optimization problem. This entry presents a novel feature selection method based on a binary gravitational search algorithm (BGSA). The proposed method, which is called modified BGSA (MBGSA), uses BGSA for performing the global search to find the best subset of features through the wrapper method. Moreover, for improving the efficiency of BGSA, mutual information (MI) feature selector under the uniform information distribution (MIFS-U) method, which works as a filter method, is integrated into BGSA as the inner optimization layer. In fact, with the computation of the relevance between each selected feature and the target class and the redundancy between the selected features (in the feature subset generated by the wrapper), MIFS-U will find more valuable features that have maximum relevance to the target class and minimum redundancy to each other. The experimental results on NSL-KDD dataset using different classifiers show that the proposed method can find better subset features and achieve higher accuracy and an improved detection rate using fewer features as compared to standard BGSA and binary particle swarm optimization (BPSO) feature selection methods.

Optimum-path forest (OPF) is a novel supervised graph-based classifier which reduces the classification problem into partitioning of vertices in a graph derived from the data samples. One of the main processes in OPF is identifying the optimum set of key samples named prototypes. This process is based on creating a minimum spanning tree on a complete weighted graph which is derived from the training samples; hence, it is much time-consuming for large-scale problems. In this study, for overcoming this limitation, the process of finding the prototypes in traditional OPF is modified by using Markov cluster (MCL) algorithm. The graph partitioning in MCL is based on finding key samples named attractors, which attract other related samples; so the obtained attractors can be selected as prototypes for generating optimum-path trees. Experiments on public benchmark datasets show that the speed of proposed modified OPF is improved considerably as compared to the traditional OPF.

In computer networks, Internet of things (IoT) is an emerging paradigm wherein smart and resource-constrained objects can connect to Internet by using a wide range of technologies. Due to the insecure nature of Internet and also wireless sensor networks (WSNs), which are the main components of IoT, implementing security mechanisms in IoT seems necessary. To deal with intrusions which may occur in IoT, a novel intrusion detection architecture model for IoT is proposed in this paper. This model is based on MapReduce approach with the aim of distributed detection. To provide multi-faceted detection (from the Internet and WSNs sides), the proposed model consists of anomaly-based and misuse-based intrusion detection agents that use supervised and unsupervised optimum-path forest model for intrusion detection. The experimental results of simulated scenarios show the superior performance of proposed method in intrusion detection for IoT.

Intrusion detection systems (IDSs) play an important role in the security of computer networks. One of the main challenges in IDSs is the high-dimensional input data analysis. Feature selection is a solution to overcoming this problem. This paper presents a hybrid feature selection method using binary gravitational search algorithm (BGSA) and mutual information (MI) for improving the efficiency of standard BGSA as a feature selection algorithm. The proposed method, called MI-BGSA, used BGSA as a wrapper-based feature selection method for performing global search. Moreover, MI approach was integrated into the BGSA, as a filter-based method, to compute the feature–feature and the feature–class mutual information with the aim of pruning the subset of features. This strategy found the features considering the least redundancy to the selected features and also the most relevance to the target class. A two-objective function based on maximizing the detection rate and minimizing the false positive rate was defined as a fitness function to control the search direction of the standard BGSA. The experimental results on the NSL-KDD dataset showed that the proposed method can reduce the feature space dramatically. Moreover, the proposed algorithm found better subset of features and achieved higher accuracy and detection rate as compared to the some standard wrapper-based and filter-based feature selection methods.