Biblio

Poisoning attack in which an adversary misleads the learning process by manipulating its training set significantly affect the performance of classifiers in security applications. This paper proposed a robust learning method which reduces the influences of attack samples on learning. The sensitivity, defined as the fluctuation of the output with small perturbation of the input, in Localized Generalization Error Model (L-GEM) is measured for each training sample. The classifier's output on attack samples may be sensitive and inaccurate since these samples are different from other untainted samples. An import score is assigned to each sample according to its localized generalization error bound. The classifier is trained using a new training set obtained by resampling the samples according to their importance scores. RBFNN is applied as the classifier in experimental evaluation. The proposed model outperforms than the traditional one under the well-known label flip poisoning attacks including nearest-first and farthest-first flips attack.

Malware has become sophisticated and organizations don't have a Plan B when standard lines of defense fail. These failures have devastating consequences for organizations, such as sensitive information being exfiltrated. A promising avenue for improving the effectiveness of behavioral-based malware detectors is to combine fast (usually not highly accurate) traditional machine learning (ML) detectors with high-accuracy, but time-consuming, deep learning (DL) models. The main idea is to place software receiving borderline classifications by traditional ML methods in an environment where uncertainty is added, while software is analyzed by time-consuming DL models. The goal of uncertainty is to rate-limit actions of potential malware during deep analysis. In this paper, we describe Chameleon, a Linux-based framework that implements this uncertain environment. Chameleon offers two environments for its OS processes: standard - for software identified as benign by traditional ML detectors - and uncertain - for software that received borderline classifications analyzed by ML methods. The uncertain environment will bring obstacles to software execution through random perturbations applied probabilistically on selected system calls. We evaluated Chameleon with 113 applications from common benchmarks and 100 malware samples for Linux. Our results show that at threshold 10%, intrusive and non-intrusive strategies caused approximately 65% of malware to fail accomplishing their tasks, while approximately 30% of the analyzed benign software to meet with various levels of disruption (crashed or hampered). We also found that I/O-bound software was three times more affected by uncertainty than CPU-bound software.

With a large number of sensors and control units in networked systems, distributed support vector machines (DSVMs) play a fundamental role in scalable and efficient multi-sensor classification and prediction tasks. However, DSVMs are vulnerable to adversaries who can modify and generate data to deceive the system to misclassification and misprediction. This work aims to design defense strategies for DSVM learner against a potential adversary. We use a game-theoretic framework to capture the conflicting interests between the DSVM learner and the attacker. The Nash equilibrium of the game allows predicting the outcome of learning algorithms in adversarial environments, and enhancing the resilience of the machine learning through dynamic distributed algorithms. We develop a secure and resilient DSVM algorithm with rejection method, and show its resiliency against adversary with numerical experiments.

The Extensible Markup Language (XML) is a complex language, and consequently, XML-based protocols are susceptible to entire classes of implicit and explicit security problems. Message formats in XML-based protocols are usually specified in XML Schema, and as a first-line defense, schema validation should reject malformed input. However, extension points in most protocol specifications break validation. Extension points are wildcards and considered best practice for loose composition, but they also enable an attacker to add unchecked content in a document, e.g., for a signature wrapping attack. This paper introduces datatyped XML visibly pushdown automata (dXVPAs) as language representation for mixed-content XML and presents an incremental learner that infers a dXVPA from example documents. The learner generalizes XML types and datatypes in terms of automaton states and transitions, and an inferred dXVPA converges to a good-enough approximation of the true language. The automaton is free from extension points and capable of stream validation, e.g., as an anomaly detector for XML-based protocols. For dealing with adversarial training data, two scenarios of poisoning are considered: a poisoning attack is either uncovered at a later time or remains hidden. Unlearning can therefore remove an identified poisoning attack from a dXVPA, and sanitization trims low-frequent states and transitions to get rid of hidden attacks. All algorithms have been evaluated in four scenarios, including a web service implemented in Apache Axis2 and Apache Rampart, where attacks have been simulated. In all scenarios, the learned automaton had zero false positives and outperformed traditional schema validation.

With the development of cyber threats on the Internet, the number of malware, especially unknown malware, is also dramatically increasing. Since all of malware cannot be analyzed by analysts, it is very important to find out new malware that should be analyzed by them. In order to cope with this issue, the existing approaches focused on malware classification using static or dynamic analysis results of malware. However, the static and the dynamic analyses themselves are also too costly and not easy to build the isolated, secure and Internet-like analysis environments such as sandbox. In this paper, we propose a lightweight malware classification method based on detection results of anti-virus software. Since the proposed method can reduce the volume of malware that should be analyzed by analysts, it can be used as a preprocess for in-depth analysis of malware. The experimental showed that the proposed method succeeded in classification of 1,000 malware samples into 187 unique groups. This means that 81% of the original malware samples do not need to analyze by analysts.

Despite widespread use of commercial anti-virus products, the number of malicious files detected on home and corporate computers continues to increase at a significant rate. Recently, anti-virus companies have started investing in machine learning solutions to augment signatures manually designed by analysts. A malicious file's determination is often represented as a hierarchical structure consisting of a type (e.g. Worm, Backdoor), a platform (e.g. Win32, Win64), a family (e.g. Rbot, Rugrat) and a family variant (e.g. A, B). While there has been substantial research in automated malware classification, the aforementioned hierarchical structure, which can provide additional information to the classification models, has been ignored. In this paper, we propose the novel idea and study the performance of employing hierarchical learning algorithms for automated classification of malicious files. To the best of our knowledge, this is the first research effort which incorporates the hierarchical structure of the malware label in its automated classification and in the security domain, in general. It is important to note that our method does not require any additional effort by analysts because they typically assign these hierarchical labels today. Our empirical results on a real world, industrial-scale malware dataset of 3.6 million files demonstrate that incorporation of the label hierarchy achieves a significant reduction of 33.1% in the binary error rate as compared to a non-hierarchical classifier which is traditionally used in such problems.

Malware classification is a critical part in the cyber-security. Traditional methodologies for the malware classification typically use static analysis and dynamic analysis to identify malware. In this paper, a malware classification methodology based on its binary image and extracting local binary pattern (LBP) features is proposed. First, malware images are reorganized into 3 by 3 grids which is mainly used to extract LBP feature. Second, the LBP is implemented on the malware images to extract features in that it is useful in pattern or texture classification. Finally, Tensorflow, a library for machine learning, is applied to classify malware images with the LBP feature. Performance comparison results among different classifiers with different image descriptors such as GIST, a spatial envelop, and the LBP demonstrate that our proposed approach outperforms others.

Detecting malicious code with exact match on collected datasets is becoming a large-scale identification problem due to the existence of new malware variants. Being able to promptly and accurately identify new attacks enables security experts to respond effectively. My proposal is to develop an automated framework for identification of unknown vulnerabilities by leveraging current neural network techniques. This has a significant and immediate value for the security field, as current anti-virus software is typically able to recognize the malware type only after its infection, and preventive measures are limited. Artificial Intelligence plays a major role in automatic malware classification: numerous machine-learning methods, both supervised and unsupervised, have been researched to try classifying malware into families based on features acquired by static and dynamic analysis. The value of automated identification is clear, as feature engineering is both a time-consuming and time-sensitive task, with new malware studied while being observed in the wild.

Nowadays, Malware has become a serious threat to the digitization of the world due to the emergence of various new and complex malware every day. Due to this, the traditional signature-based methods for detection of malware effectively becomes an obsolete method. The efficiency of the machine learning model in context to the detection of malware files has been proved by different researches and studies. In this paper, a framework has been developed to detect and classify different files (e.g exe, pdf, php, etc.) as benign and malicious using two level classifier namely, Macro (for detection of malware) and Micro (for classification of malware files as a Trojan, Spyware, Adware, etc.). Cuckoo Sandbox is used for generating static and dynamic analysis report by executing files in the virtual environment. In addition, a novel model is developed for extracting features based on static, behavioral and network analysis using analysis report generated by the Cuckoo Sandbox. Weka Framework is used to develop machine learning models by using training datasets. The experimental results using proposed framework shows high detection rate with an accuracy of 100% using J48 Decision tree model, 99% using SMO (Sequential Minimal Optimization) and 97% using Random Forest tree. It also shows effective classification rate with accuracy 100% using J48 Decision tree, 91% using SMO and 66% using Random Forest tree. These results are used for detecting and classifying unknown files as benign or malicious.

Outside the highly publicized victories in the game of Go, there have been numerous successful applications of deep learning in the fields of information retrieval, computer vision, and speech recognition. In cybersecurity, an increasing number of companies have begun exploring the use of deep learning (DL) in a variety of security tasks with malware detection among the more popular. These companies claim that deep neural networks (DNNs) could help turn the tide in the war against malware infection. However, DNNs are vulnerable to adversarial samples, a shortcoming that plagues most, if not all, statistical and machine learning models. Recent research has demonstrated that those with malicious intent can easily circumvent deep learning-powered malware detection by exploiting this weakness. To address this problem, previous work developed defense mechanisms that are based on augmenting training data or enhancing model complexity. However, after analyzing DNN susceptibility to adversarial samples, we discover that the current defense mechanisms are limited and, more importantly, cannot provide theoretical guarantees of robustness against adversarial sampled-based attacks. As such, we propose a new adversary resistant technique that obstructs attackers from constructing impactful adversarial samples by randomly nullifying features within data vectors. Our proposed technique is evaluated on a real world dataset with 14,679 malware variants and 17,399 benign programs. We theoretically validate the robustness of our technique, and empirically show that our technique significantly boosts DNN robustness to adversarial samples while maintaining high accuracy in classification. To demonstrate the general applicability of our proposed method, we also conduct experiments using the MNIST and CIFAR-10 datasets, widely used in image recognition research.

In an attempt to preserve the structural information in malware binaries during feature extraction, function call graph-based features have been used in various research works in malware classification. However, the approach usually employed when performing classification on these graphs, is based on computing graph similarity using computationally intensive techniques. Due to this, much of the previous work in this area incurred large performance overhead and does not scale well. In this paper, we propose a linear time function call graph (FCG) vector representation based on function clustering that has significant performance gains in addition to improved classification accuracy. We also show how this representation can enable using graph features together with other non-graph features.

Distinguishing and classifying different types of malware is important to better understanding how they can infect computers and devices, the threat level they pose and how to protect against them. In this paper, a system for classifying malware programs is presented. The paper describes the architecture of the system and assesses its performance on a publicly available database (provided by Microsoft for the Microsoft Malware Classification Challenge BIG2015) to serve as a benchmark for future research efforts. First, the malicious programs are preprocessed such that they are visualized as gray scale images. We then make use of an architecture comprised of multiple layers (multiple levels of encoding) to carry out the classification process of those images/programs. We compare the performance of this approach against traditional machine learning and pattern recognition algorithms. Our experimental results show that the deep learning architecture yields a boost in performance over those conventional/standard algorithms. A hold-out validation analysis using the superior architecture shows an accuracy in the order of 99.15%.

Anti-virus vendors receive hundreds of thousands of malware to be analysed each day. Some are new malware while others are variations or evolutions of existing malware. Because analyzing each malware sample by hand is impossible, automated techniques to analyse and categorize incoming samples are needed. In this work, we explore various machine learning features extracted from malware samples through static analysis for classification of malware binaries into already known malware families. We present a new feature based on control statement shingling that has a comparable accuracy to ordinary opcode n-gram based features while requiring smaller dimensions. This, in turn, results in a shorter training time.

Malware technology makes it difficult for malware analyst to detect same malware files with different obfuscation technique. In this paper we are trying to tackle that problem by analyzing the sequence of system call from an executable file. Malware files which actually are the same should have almost identical or at least a similar sequence of system calls. In this paper, we are going to create a model for each malware class consists of malwares from different families based on its sequence of system calls. Method/algorithm that's used in this paper is profile hidden markov model which is a very well-known tool in the biological informatics field for comparing DNA and protein sequences. Malware classes that we are going to build are trojan and worm class. Accuracy for these classes are pretty high, it's above 90% with also a high false positive rate around 37%.

Malware writers often develop malware with automated measures, so the number of malware has increased dramatically. Automated measures tend to repeatedly use significant modules, which form the basis for identifying malware variants and discriminating malware families. Thus, we propose a novel visualization analysis method for researching malware similarity. This method converts malicious Windows Portable Executable (PE) files into local entropy images for observing internal features of malware, and then normalizes local entropy images into entropy pixel images for malware classification. We take advantage of the Jaccard index to measure similarities between entropy pixel images and the k-Nearest Neighbor (kNN) classification algorithm to assign entropy pixel images to different malware families. Preliminary experimental results show that our visualization method can discriminate malware families effectively.

Comparison of two binary files has many practical applications: the ability to detect programmatic changes between two versions, the ability to find old versions of statically linked libraries to prevent the use of well-known bugs, malware analysis, etc. In this article, a framework for comparison of binary files is presented. Framework uses IdaPro [1] disassembler and Binnavi [2] platform to recover structure of the target program and represent it as a call graph (CG). A program dependence graph (PDG) corresponds to each vertex of the CG. The proposed comparison algorithm consists of two main stages. At the first stage, several heuristics are applied to find the exact matches. Two functions are matched if at least one of the calculated heuristics is the same and unique in both binaries. At the second stage, backward and forward slicing is applied on matched vertices of CG to find further matches. According to empiric results heuristic method is effective and has high matching quality for unchanged or slightly modified functions. As a contradiction, to match heavily modified functions, binary code clone detection is used and it is based on finding maximum common subgraph for pair of PDGs. To achieve high performance on extensive binaries, the whole matching process is parallelized. The framework is tested on the number of real world libraries, such as python, openssh, openssl, libxml2, rsync, php, etc. Results show that in most cases more than 95% functions are truly matched. The tool is scalable due to parallelization of functions matching process and generation of PDGs and CGs.

The heterogeneous SIS model for virus spread in any finite size graph characterizes the influence of factors of SIS model and could be analyzed by the extended N-Intertwined model introduced in [1]. We specifically focus on the heterogeneous virus spread in the star network in this paper. The epidemic threshold and the average meta-stable state fraction of infected nodes are derived for virus spread in the star network. Our results illustrate the effect of the factors of SIS model on the steady state infection.

Malicious applications have become increasingly numerous. This demands adaptive, learning-based techniques for constructing malware detection engines, instead of the traditional manual-based strategies. Prior work in learning-based malware detection engines primarily focuses on dynamic trace analysis and byte-level n-grams. Our approach in this paper differs in that we use compiler intermediate representations, i.e., the callgraph representation of binaries. Using graph-based program representations for learning provides structure of the program, which can be used to learn more advanced patterns. We use the Shortest Path Graph Kernel (SPGK) to identify similarities between call graphs extracted from binaries. The output similarity matrix is fed into a Support Vector Machine (SVM) algorithm to construct highly-accurate models to predict whether a binary is malicious or not. However, SPGK is computationally expensive due to the size of the input graphs. Therefore, we evaluate different parallelization methods for CPUs and GPUs to speed up this kernel, allowing us to continuously construct up-to-date models in a timely manner. Our hybrid implementation, which leverages both CPU and GPU, yields the best performance, achieving up to a 14.2x improvement over our already optimized OpenMP version. We compared our generated graph-based models to previously state-of-the-art feature vector 2-gram and 3-gram models on a dataset consisting of over 22,000 binaries. We show that our classification accuracy using graphs is over 19% higher than either n-gram model and gives a false positive rate (FPR) of less than 0.1%. We are also able to consider large call graphs and dataset sizes because of the reduced execution time of our parallelized SPGK implementation.

Botnets are a growing threat to the security of data and services on a global level. They exploit vulnerabilities in networks and host machines to harvest sensitive information, or make use of network resources such as memory or bandwidth in cyber-crime campaigns. Bot programs by nature are largely automated and systematic, and this is often used to detect them. In this paper, we extend upon existing work in this area by proposing a network event correlation method to produce graphs of flows generated by botnets, outlining the implementation and functionality of this approach. We also show how this method can be combined with statistical flow-based analysis to provide a descriptive chain of events, and test on public datasets with an overall success rate of 94.1%.

As a vital component of variety cyber attacks, malicious domain detection becomes a hot topic for cyber security. Several recent techniques are proposed to identify malicious domains through analysis of DNS data because much of global information in DNS data which cannot be affected by the attackers. The attackers always recycle resources, so they frequently change the domain - IP resolutions and create new domains to avoid detection. Therefore, multiple malicious domains are hosted by the same IPs and multiple IPs also host same malicious domains in simultaneously, which create intrinsic association among them. Hence, using the labeled domains which can be traced back from queries history of all domains to verify and figure out the association of them all. Graphs seem the best candidate to represent for this relationship and there are many algorithms developed on graph with high performance. A graph-based interface can be developed and transformed to the graph mining task of inferring graph node's reputation scores using improvements of the belief propagation algorithm. Then higher reputation scores the nodes reveal, the more malicious probabilities they infer. For demonstration, this paper proposes a malicious domain detection technique and evaluates on a real-world dataset. The dataset is collected from DNS data servers which will be used for building a DNS graph. The proposed technique achieves high performance in accuracy rates over 98.3%, precision and recall rates as: 99.1%, 98.6%. Especially, with a small set of labeled domains (legitimate and malicious domains), the technique can discover a large set of potential malicious domains. The results indicate that the method is strongly effective in detecting malicious domains.

We present a binary static analysis approach to detect intelligent electronic device (IED) malware based on the time requirements of electrical substations. We explore graph theory techniques to model the timing performance of an IED executable. Timing performance is subsequently used as a metric for IED malware detection. More specifically, we perform a series of steps to reduce a part of the IED malware detection problem into a classical problem of graph theory, namely finding single-source shortest paths on a weighted directed acyclic graph (DAG). Shortest paths represent execution flows that take the longest time to compute. Their clock cycles are examined to determine if they violate the real-time nature of substation monitoring and control, in which case IED malware detection is attained. We did this work with particular reference to implementations of protection and control algorithms that use the IEC 61850 standard for substation data representation and network communication. We tested our approach against IED exploits and malware, network scanning code, and numerous malware samples involved in recent ICS malware campaigns.

Due to the unavailability of signatures for previously unknown malware, non-signature malware detection schemes typically rely on analyzing program behavior. Prior behavior based non-signature malware detection schemes are either easily evadable by obfuscation or are very inefficient in terms of storage space and detection time. In this paper, we propose GZero, a graph theoretic approach fast and accurate non-signature malware detection at end hosts. GZero it is effective while being efficient in terms of both storage space and detection time. We conducted experiments on a large set of both benign software and malware. Our results show that GZero achieves more than 99% detection rate and a false positive rate of less than 1%, with less than 1 second of average scan time per program and is relatively robust to obfuscation attacks. Due to its low overheads, GZero can complement existing malware detection solutions at end hosts.

With growing popularity of Android, it's attack surface has also increased. Prevalence of third party android marketplaces gives attackers an opportunity to plant their malicious apps in the mobile eco-system. To evade signature based detection, attackers often transform their malware, for instance, by introducing code level changes. In this paper we propose a lightweight static Permission Flow Graph (PFG) based approach to detect malware even when they have been transformed (obfuscated). A number of techniques based on behavioral analysis have also been proposed in the past; how-ever our interest lies in leveraging the permission framework alone to detect malware variants and transformations without considering behavioral aspects of a malware. Our proposed approach constructs Permission Flow Graph (PFG) for an Android App. Transformations performed at code level, often result in changing control flow, however, most of the time, the permission flow remains invariant. As a consequences, PFGs of transformed malware and non-transformed malware remain structurally similar as shown in this paper using state-of-the-art graph similarity algorithm. Furthermore, we propose graph based similarity metrics at both edge level and vertex level in order to bring forth the structural similarity of the two PFGs being compared. We validate our proposed methodology through machine learning algorithms. Results prove that our approach is successfully able to group together Android malware and its variants (transformations) together in the same cluster. Further, we demonstrate that our proposed approach is able to detect transformed malware with a detection accuracy of 98.26%, thereby ensuring that malicious Apps can be detected even after transformations.

The analysis of multiple Android malware families indicates malware instances within a common malware family always have similar call graph structures. Based on the isomorphism of sensitive API call graph, we propose a method which is used to construct malware family features via combining static analysis approach with graph similarity metric. The experiment is performed on a malware dataset which contains 1326 malware samples from 16 different malware families. The result shows that the method can differentiate distinct malware family features and divide suspect malware samples into corresponding families with a high accuracy of 96.77% overall and even defend a certain extent of obfuscation.

With the advancement of unmanned aerial vehicles (UAV), 3D wireless mesh networks will play a crucial role in next generation mission critical wireless networks. Along with providing coverage over difficult terrain, it provides better spectral utilization through 3D spatial reuse. However, being a wireless network, 3D meshes are vulnerable to jamming/disruptive attacks. A jammer can disrupt the communication, as well as control of the network by intelligently causing interference to a set of nodes. This paper presents a distributed mechanism of avoiding jamming attacks by means of 3D spatial filtering where adaptive beam nulling is used to keep the jammer in null region in order to bypass jamming. Kalman filter based tracking mechanism is used to estimate the most likely trajectory of the jammer from noisy observation of the jammer's position. A beam null border is determined by calculating confidence region of jammer's current and next position estimates. An optimization goal is presented to calculate optimal beam null that minimizes the number of deactivated links while maximizing the higher value of confidence for keeping the jammer inside the null. The survivability of a 3D mesh network with a mobile jammer is studied through simulation that validates an 96.65% reduction in the number of jammed nodes.