NISL Researcher

Hamid Bostani

PhD Candidate in the Digital Security Group
Institute for Computing and Information Sciences
Radboud University, The Netherlands

I am a well-organized, hard-working, and ambitious person who fights to achieve his dreams. I believe that my strong enthusiasm as well as my intrinsic capabilities in problem-solving, along with my proper academic and technical experiences, are convincing signs that can promise me to gain high achievements in computer science.

About me

My primary research interest is investigating AI in systems security because I strongly feel that AI is a promising problem solver that can obtain satisfying solutions for complex problems. Indeed, the main area of my research is at the intersection of machine learning and systems security by concentrating on intrusion & malware detection.

My current research focus is adversarial machine learning in the area of malware detection, which I am working on under the supervision of Dr. Veelasha Moonsamy and Prof. Erik Poll.

Previously, I was a Research Assistant at Islamic Azad University (IAU), South Tehran Branch who was working on AI, particularly scalable machine learning. My considerable background in AI began on my master's project while I was trying to present an AI-based intrusion detection system for IoT that could detect various security threats. During this time, I have succeeded to present several novelties in machine learning, IoT security, and feature selection that brought various honors and awards for me.

From September 2012 till September 2020, I was working in the National Organization of Educational Testing, Tehran, Iran as a Full-Stack Developer. It was a useful industrial career that made me an experienced software developer.

I received my BSc and MSc degrees in Computer Engineering from IAU, Shiraz and South Tehran Branches, Iran in 2008 and 2015, respectively.


20 December 2023, My manuscript entitled "EvadeDroid: A practical evasion attack on machine learning for black-box Android malware detection, " has been accepted by the Computers & Security Journal.
28 Ocotober 2023,  The paper I co-authored, titled "Targeted and Troublesome: Tracking and Advertising on Children's Websites," has been accepted at the 45th IEEE Symposium on Security and Privacy (IEEE S&P'24).
2 Ocotober 2023,  I will visit King's College London and University College London for 6 months and work on Adversarial Training under the supervision of Dr. Fabio Pierazzi and Prof. Lorenzo Cavallaro.
20 April 2023, I attend ICT.OPEN2023, Utrecht, The Netherlands to present my poster entitled "Improving Robustness of Machine Learning-based Android Malware Detection against Realizable Adversarial Examples.
1 August 2022, I attend the Summer School on Privacy-Preserving Machine Learning (PPML) organized by the department of Computer Science at ITU Copenhagen, Denmark.

14 January 2022, My book chapter entitled "Hybrid and modified OPFs for intrusion detection systems and large-scale problems, " is published as Chapter 5 of Elsevier's book, Optimum-Path Forest: Theory, Algorithms, and Applications.

1 October 2020, I have been employed as a PhD Candidate in the Digital Security Group at Radboud University.

23 August 2020, My manuscript entitled "A Strong Coreset Algorithm to Accelerate OPF as a Graph-based Machine Learning in Large-Scale Problems, " has been accepted by the Information Sciences Journal with a minor revision.


Security and Privacy of Smart Mobile Devices
July 2020 - Present
Over the past decade, the digital life of most people has moved from stationary devices such as desktop computers to computationally powerful mobile devices, such as smartphones, tablets or smartwatches. Not only do these devices bring a new level of convenience for users, they are also much more vulnerable to attacks that compromise the confidentiality of users' private data - as revealed by whistleblower Edward Snowden's global surveillance disclosures. One might think that similar attacks existed before, but mobile smart devices over many new attack vectors that are not known from stationary desktop PCs. For example, a new class of attack vector results from the plethora of sensors these devices carry, such as GPS and motion sensors. While some of these attacks can be thwarted by sensible use of strong cryptography - for instance, there now exist secure text-messenger apps that allow users to encrypt their communication - strong cryptography alone is not able to protect against attacks that exploit side channels. Side channels allow the exploitation of information emanating from the physical device, such as electromagnetic radiation and power consumption. We are investigating the different classes of attacks on smart mobile devices through the analysis of hardware and software requirements and design countermeasures using novel and creative ideas by studying the problem from different angles. Our findings will then be applied to other problem areas such as the Internet of Things. I am working on this project as PhD Candidate in the Digital Security group, Radboud University.
Developing a New generation of Optimum-path Forest (OPF) as a Graph-based Machine Mearning in order to Achieve an Efficient Pattern Recognition Tool for Using on Massive Datasets
March 2017 - November 2019
Over the past decades, machine learning as the most efficient tool for pattern recognition has been the subject of many studies. Optimum-path forest (OPF) is an outstanding graph-based machine learning method that reduces the pattern recognition problems into the partitioning of the graphs derived from the input data sets. OPF as a natural multi-classifier is a fast, simple, and parameter-independent machine that supports partial overlapping among the classes. Howover, this machine is an effective machine learning algorithm just for a reasonable size of the input data sets. It seems that keeping a small sketch (or synopsis) of some data that contains an approximation of the original data can be a proper solution for the high computational complexity of OPF over the massive data sets. Coreset is a special data summarization method that can be used for holding the small sketch of massive data sets. Therefore, to cope with high computational complexity of OPF over massive data sets, this project tries to use the idea of coreset in the context of OPF to provide a scalable OPF. This work is supported in part by Iran National Science Foundation grant No. 96010151.
Intrusion Detection and Identification of Attacks on the Internet of Things (IoT) Using a Combination of Machine Learning Methods
May 2014 - September 2015
The Internet of Things (IoT) is a worldwide network including all identifiable heterogeneous objects around us such as smartphones, laptops or smart sensors that can connect to the Internet by using a wide range of technologies. IoT is able to provide accessibility to the Internet for all physical objects since it is a hybrid network of the Internet and diverse networks with heterogeneous nodes. Generally, due to the insecure nature of the Internet as well as Wireless Sensor Networks, which are the main components of IoT, implementing security mechanisms in IoT seems necessary. To deal with intrusions that may occur in IoT, a novel multi-faceted intrusion detection system is proposed in this thesis which can detect both cyber-attacks and insider-attacks of IoT. This project is my master thesis which was done under supervision of professor Mansour Sheikhan.

Honors & Awards

  Fully-Funded Fellowship, Radboud University (2020)
  Best Employee Award, National Organization of Educational Testing (2019)
  Certificate of Appreciation, National Organization of Educational Testing (2018)
  Research Funding, Iran National Science Foundation (2018)
  Outstanding Researcher Award, Islamic Azad University, South Tehran Branch (2017)
  Best Thesis Award, The 5th Research, Scientific & Technological National Festival of Islamic Azad University (2017)
  Certificate of Appreciation, Islamic Azad University, South Tehran Branch (2017)
  Outstanding Paper Award, The 2nd International Conference on Signal Processing and Intelligent Systems (2016)
  Selected Paper, The 8th International Symposium on Telecommunication (2016)


EvadeDroid: A practical evasion attack on machine learning for black-box Android malware detection
H. Bostani and V. Moonsamy
Computers & Security, vol. 139, pp. 1–18, 2024
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.

Chapter 5 - Hybrid and modified OPFs for intrusion detection systems and large-scale problems
M. Sheikhan, H. Bostani
In Optimum-Path Forest, pp. 109-136, Academic Press, Elsevier, 2022.
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.

A Strong Coreset Algorithm to Accelerate OPF as a Graph-based Machine Learning in Large-Scale Problems
H. Bostani, M. Sheikhan, B. Mahboobi
Information Sciences, vol. 555, pp. 424–441, 2021
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.

Developing a Fast Supervised Optimum-path Forest Based on Coreset
H. Bostani, M. Sheikhan, B. Mahboobi
In Proc. 19th International Symposium on Artificial Intelligence and Signal Processing (AISP’2017), pp. 172-177, 2017

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.

Hybrid of Anomaly-Based and Specification-Based IDS for Internet of Things Using Unsupervised OPF based on Map-Reduce Approach
H. Bostani, M. Sheikhan
Computer Communications, vol. 98, pp. 52-71, 2017

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.

Modifying Supervised Optimum-Path Forest in Intrusion Detection Systems Using Social Network Approaches and Unsupervised Learning
H. Bostani, M. Sheikhan
Pattern Recognition, vol. 62, pp. 56-72, 2017

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.

A Security Mechanism for Detecting Intrusions in Internet of Things Using Selected Features Based on MI-BGSA
M. Sheikhan, H. Bostani
International Journal of Information & Communication Technology Research, vol. 9, no. 2, pp. 53-62, 2017

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.

Binary Gravitational Search Algorithm (BGSA): Improved Efficiency
M. Sheikhan, H. Bostani
Encyclopedia of Information Assurance, 2016

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.

Modification of Optimum-Path Forest using Markov Cluster Process Algorithm
H. Bostani, M. Sheikhan
In Proc. 2nd International Conference on Signal Processing and Intelligent Systems (ICSPIS’2016), pp. 1-5, 2016 (Winner of the Outstanding Paper Award)

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.

A Hybrid Intrusion Detection Architecture for Internet of Things
M. Sheikhan, H. Bostani
In Proc. 8th International Symposium on Telecommunication (IST’2016), pp. 601-606, 2016 (Selected as one of the Best Paper)

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.

Hybrid of Binary Gravitational Search Algorithm and Mutual Information for Feature Selection in Intrusion Detection Systems
H. Bostani, M. Sheikhan
Soft Computing, vol. 21, no. 9, pp. 2307-2324, 2017

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.


Professional National Certifications
  SQL Server Query Tuning and Optimization, Faratar As Danesh Institute, Tehran, Iran (2018)
  SQL Server 2016 – Design & Implementation, Faratar As Danesh Institute, Tehran, Iran (2018)
  Professional SCRUM Master, Faratar As Danesh Institute, Tehran, Iran (2016)
  MCSD Web Pack 2012, Kahkeshane Noor Institute, Tehran, Iran (2016)
  ETL (SSIS) and Data Mining (SSAS) 2012, Faratar As Danesh Institute, Tehran, Iran (2014)
  Data Warehousing & OLAP using SSAS 2012, Faratar As Danesh Institute, Tehran, Iran (2014)
  Win Application (C# & intro ADO.NET), South Industrial Management Institute, Shiraz, Iran (2008)

Computer Knowledge
  Programming and Scripting: C/C++, C#, Java, HTML & CSS, MATLAB, Python
  Tools and IDEs: MATLAB Optimization, Fuzzy, and Neural Net Tools, Microsoft Visual Studio, Microsoft SQL Server, Visual Paradigm, Microsoft Office, Azure Boards
  Software Development Technologies: C#.Net Windows Form, ASP.Net Web Form, APS.Net MVC, Javascritp, JQuery & AngularJS, ADO.NET Entity Framework, Database (programming), Java 2 Platform Micro Edition (J2ME), Microsoft BI Technologies (Analysis, Integrated, and Reporting Services)
  Software Development Methodologies: RUP, EUP, SCRUM
  OS: Windows