Publications 

The growth of the Internet of Things (IoT), driven by Wireless Sensor Networks (WSNs) and Low-Power and Lossy Networks (LLNs), has raised significant security concerns, particularly regarding sinkhole attacks. These attacks compromise data integrity by redirecting traffic through malicious nodes. This paper reviews existing detection methods for sinkhole attacks in IoT and WSNs, analyzing their effectiveness, limitations, and current research gaps. Key challenges include the lack of benchmark datasets, limited real-world validation, and issues with scalability and protocol dependence. The study highlights future research directions to enhance the security and resilience of IoT systems. 

Check our work here.

The proliferated smart TV market has sparked a race among the tech giants to capture market share, with Google aggressively pursuing this domain through partnerships with third-party smart TV manufacturers. However, this expansion raises critical concerns regarding security vulnerabilities and the potential breakdown of Google’s Trustworthy Artificial Intelligence Implementation (TAII) standards. Despite the significant focus on software security, there is a lack of research on the security and privacy challenges associated with smart TV hardware. In this paper, we present a case study exploring the security vulnerabilities present in TCL Smart TVs integrated with Google Cloud services. Our case study result identifies significant risks of consumer data breaches and exposure of personal information. Our findings show significant gaps in following the TAII principles of privacy, security, safety, and transparency. 

Check our work here.

The extraction of keywords is a critical task in natural language processing and information retrieval. It has become increasingly important in a wide range of applications, from search engines and e-commerce platforms to news and social media analysis. However, the evaluation of keyword extraction methods remains a challenging task due to the diverse range of data types and contexts used, as well as the complexity of the methodologies and techniques involved. In this regard, this study reviews the prior surveys on keyword extraction methods to comprehend the fundamental principles, difficulties in keyword extraction, and benchmark datasets. The reliability of evaluation techniques and an examination of their flaws remain two of the largest problems. Hence, this study includes the literature that performed a comparative analysis of popular keyword extraction methods. Furthermore, in this paper, we present a comparative evaluation of open-source unsupervised keyword extraction tools, analyzing their performance across a range of data types and under different testing conditions. The experimental analysis shows that, in terms of f-score, KPminer performs most consistently for different text lengths while KeyBERT(mmr) outperforms other tools. Considering the execution time, RAKE and YAKE are the fastest tools. Though graph-based tools tend not to perform well on long text, TopicRank and MultipartiteRank perform very well on long text as they use topics as nodes of the graph, which is another finding of this study. By highlighting the key factors that influence the performance of keyword extraction tools, our analysis contributes to directing the reader in selecting suitable keyword extraction tools for different applications. 

Check our work here.

Study groups are crucial in education to foster collaboration and motivation. Remote students, at greater risk of isolation, often miss this peer-driven environment. This paper presents Buenas, an Extended Reality (XR) system enabling remote students to participate in on-campus study groups seamlessly. Local students interact with projected representations of remote peers seated at a conference table, while remote students use XR headsets to integrate the group into their physical surroundings. A user study comparing Buenas to conventional videoconferencing showed significant improvements in engagement, presence, connectedness, and rapport through both objective and subjective measures. 

Check our work here

In the era of data-driven decision making, cities and communities are increasingly seeking ways to effectively gather insights from public feedback and comments to shape their research and development initiatives. Town hall community meetings serve as a valuable platform for citizens to express their opinions, concerns, and ideas about various aspects of city life. In this study, we aim to explore the effectiveness of different keyword extraction tools and similarity matching algorithms in matching town hall community comments with city strategic plans and current research opportunities. We employ KPMiner, TopicRank, MultipartiteRank, and KeyBERT for keyword extraction, and evaluate the performance of cosine similarity, word embedding similarity, and BERT-based similarity for matching the extracted keywords. By combining these techniques, we aim to bridge the gap between community feedback and research initiatives, enabling data-driven decision-making in urban development. Our findings will provide valuable insights for more inclusive and informed strategies, ensuring that citizen opinions and concerns are effectively incorporated into city planning and development efforts. 

Check our work here

Urban governance is vital for efficiently managing cities, promoting sustainable development, and improving quality of life for residents. In the realm of urban governance, the San Antonio Research Partnership Portal stands as a groundbreaking initiative, fostering collaboration between diverse city entities and leveraging innovative smart applications. In this paper, we will focus on its ability to facilitate strategic alignment among city departments, public feedback integration, and streamlined collaboration with academic institutions. Through technical insights and real-world case studies, this paper underscores the portal’s role in enhancing municipal responsiveness, improving decision-making processes, and exemplifying the potential of smart applications utilizing artificial intelligence for fostering effective city management and community engagement. 

Check our work here

Finding research professionals and collaborators to address community problems continues to be a significant barrier for many local government agencies. Research collaboration between researchers from universities, industries and local government agencies can be tremendously useful to all organizations. San Antonio Research Partnership Portal is a collaborative initiative to bring researchers and local government agencies in one place to solve community concerns. In this paper, we investigate the performance of popular keyword extraction tools by measuring the effectiveness of identifying the keywords from research opportunities. The extracted keywords are used in an automated process for San Antonio Research Partnership Portal to match academic researchers with corresponding research opportunities. 

Check our work here.

Research collaboration between academic researchers from universities, organizations and government local city departments can be tremendously useful to all institutions, but these collaborations involve a wide range of skill sets, making them difficult to establish and manage. For many local government city departments, finding research professionals and collaborators to solve community problems remains a big challenge. The information of research opportunities is either posted on the individual city department website or researchers are approached through a personal relationship with city department officials. As a result, a researcher interested in working with the city department would have to either navigate various websites or try to build a personal contact with city department officials. In this paper, we will look at the relevance of community research partnerships as well as the barriers that prevent them. We will also demonstrate the development of, Research Partnership Portal, a collaborative platform for academic researchers, organizations, and government city departments in San Antonio. This portal will assist academic researchers and organizations in collaborating with government city departments and using current administrative data to produce effective answers to community concerns.

Check our work here.

The core part of the computer operating system that plays an important role in managing computer resources is the kernel. One of the most elusive types of malware in recent times that pose significant security threats on the computer operating system kernel is the kernel-level rootkit. The kernel-level rootkit can hide its presence and malicious activities by modifying the kernel control flow, by hooking in the kernel space, or by manipulating the kernel objects. As kernel-level rootkit changes the kernel, it is difficult for user-level security tools to detect the kernel-level rootkit. In the past few years, researchers have proposed and experimented with many detection systems to detect the evolving kernel-level rootkit. A learning-based detection is an excellent approach to automatically detect known and unknown attacks with high accuracy. In this paper, we have reviewed the prior learning-based approaches in the literature that detect the kernel-level rootkit. We have also discussed the strengths and weaknesses of prior learning-based detection approaches against the kernel-level rootkit. The paper ends with open issues, challenges, and future research direction for the kernel-level rootkit detection.

Check our work here.

The container-based cloud computing service is increasingly adopted by many service providers for its efficiency and flexibility. Containers isolated by namespaces share OS kernel. When the kernel-level rootkits exploit vulnerabilities existing in kernel, the namespace can be invalidated leading to critical security incidents. Even though many traditional approaches have been made to detect kernel-level rootkits, it is hard to detect new attacks conducted in the new environment such as container-based cloud computing system. In this paper, we show some possible attack scenarios by kernel-level rootkits exploiting kernel namespaces and suggest key features that can be used to train machine learning and neural network models. 

Check our work here

The core part of the operating system is the kernel, and it plays an important role in managing critical data structure resources for correct operations. The kernel-level rootkits are the most elusive type of malware that can modify the running OS kernel in order to hide its presence and perform many malicious activities such as process hiding, module hiding, network communication hiding, and many more. In the past years, many approaches have been proposed to detect kernel-level rootkit. Still, it is challenging to detect new attacks and properly categorize the kernel-level rootkits. Memory forensic approaches showed efficient results with the limitation against transient attacks. Cross-view-based and integrity monitoring-based approaches have their own weaknesses. A learning-based detection approach is an excellent way to solve these problems. In this paper, we give an insight into the kernel-level rootkit characteristic features and how the features can be represented to train learning-based models in order to detect known and unknown attacks. Our feature set combined the memory forensic, cross-view, and integrity features to train learning-based detection models. We also suggest useful tools that can be used to collect the characteristics features of the kernel-level rootkit.

Check our work here

With the evolving artificial intelligence technology, the chatbots are becoming smarter and faster lately. Chatbots are typically available round the clock providing continuous support and services. A chatbot or a conversational agent is a program or software that can communicate using natural language with humans. The challenge of developing an intelligent chatbot still exists ever since the onset of artificial intelligence. The functionality of chatbots can range from business oriented short conversations to healthcare intervention based longer conversations. However, the primary role that the chatbots have to play is in understanding human utterances in order to respond appropriately. To that end, there is an increased emergence of Natural Language Understanding (NLU) engines by popular cloud service providers. The NLU services identify entities and intents from the user utterances provided as input. Thus, in order to integrate such understanding to a chatbot, this paper presents a study on existing major NLU platforms. Then, we present a case study chatbot integrated with a Google DialogFlow NLU service and discuss the intent recognition performance.

Check our work here

Bangladesh, a subtropical country located not very far from the equator, has certain geographical advantages in solar energy generation. But, as the solar energy generation is not as robust as other forms of energy, the solar systems must be optimally installed to reap every physical benefits possible. One very important factor in this matter is the tilt angle of the solar panel. This paper focuses on finding the optimum tilt angle for a solar panel in a particular region. The paper finds the optimum tilt angles for various divisions of Bangladesh using two methods: one optimum tilt angle for a whole year, and two tilt angles for two halves of a year. The paper demonstrates the benefit of using two seasonal tilt angles instead of using just one tilt angle throughout the year. The paper additionally checks the extent of benefit that might occur from using more complex systems involving more tilt angles. 

Check our work here