Sheikh Rabiul Islam

The rapid deployment of Artificial Intelligence (AI) in high-stakes domains necessitates robust approaches to transparency, fairness, and trustworthiness. Current advancements in AI performance are outpacing our understanding and ability to govern these systems. This special issue presents research addressing explainability, fairness, and trust as interconnected socio-technical challenges. Accepted papers demonstrate novel techniques for revealing hidden model dependencies, aligning explanations with domain expertise, rigorously benchmarking model classes for explanation robustness, and refining methods for measuring interpretability. We synthesize these contributions, situate them within current policy and standardization (EU AI Act; ¹ NIST AI RMF; ^{2 , 3} ISO/IEC 23894 and 42001 ^{4 , 5} ), and connect them to emerging evaluation science in XAI (e.g., BEExAI, ⁹ Saliency-Bench, ¹⁰ F-Fidelity ¹¹ ). Finally, we outline a forward-looking agenda emphasizing multi-aspect evaluation, context-sensitive trust, and the development of governance-ready AI systems.

With the widespread availability of adware masquerading as useful apps, there is an increasing need for robust security measures to identify adware. The identification of adware as a malware is a challenging task, as it often appears benign despite its malicious intent in the background. In this study, we propose an approach to classify adware on Android devices using data from multiple modalities. The focus is particularly on the classification of Airpush and Dowgin adware. Our proposed method uses both tabular and grayscale image data, and a feedforward neural network architecture to build a multimodal deep learning model that achieves a 95% prediction accuracy. Additionally, we incorporate Explainable AI (XAI) to enhance the interpretability of the results. The efficiency of our proposed approach is showcased through its ability to classify adware instances in an explainable manner, highlighting its significance not only in adware classification but also in fortifying against the evolving challenges posed by adware.

Despite considerable progress in malicious software forensics, the challenge of accurate attribution, formulation of appropriate response and mitigation strategies, and ensuring the interpretability of deep learning methods persists. While being less flexible and robust to noise compared to deep learning models, Knowledge Graphs are natively developed to be explainable and are a promising solution for exploring new features and relations, and enhancing understandability of decisions. In this work, we aim to develop an explainable malware classifier which can classify PE executable as malign or benign, by infusing external knowledge using Knowledge Graph (KG). We enrich our Knowledge Graph using MITRE Attack ontology (i.e., domain knowledge) and EMBER dataset and utilize Graph2Vec algorithm to embed KG knowledge into our classifier. We found that our classifier yields satisfactory results while maintaining a high level of explainability.

As an increasing number of Artificial Intelligence (AI) systems are ingrained in our day-to-day lives, it is crucial that they are fair and trustworthy. Unfortunately, this is often not the case for predictive policing systems, where there is evidence of bias towards age as well as race and sex leading to many people being mistakenly labeled as likely to be involved in a crime. In a system that already is under criticism for its unjust treatment of minority groups, it is crucial to find ways to mitigate this negative trend. In this work, we explored and evaluated the infusion of domain knowledge in the predictive policing system to minimize the prevailing fairness issues. The experimental results demonstrate an increase in fairness across all of the metrics for all of the protected classes bringing more trust into the predictive policing system by reducing the unfair policing of people.

The COVID-19 pandemic has affected millions of people worldwide with severe health, economic, social, and political implications. Healthcare Policy Makers (HPMs) and medical experts are at the core of responding to this continuously evolving pandemic situation and are working hard to contain the spread and severity of this relatively unknown virus. Biomedical researchers are continually discovering new information about this virus and communicating the findings through scientific articles. As such, it is crucial for HPMs and funding agencies to monitor the COVID-19 research trend globally on a regular basis. However, given the influx of biomedical research articles, monitoring COVID-19 research trends has become more challenging than ever, especially when HPMs want on-demand guided search techniques with a set of topics of interest in mind. Unfortunately, existing topic trend modeling techniques are unable to serve this purpose as (1) traditional topic models are unsupervised, and (2) HPMs in different regions may have different topics of interest that they want to track. To address this problem, we introduce a novel computational task in this article calledAd-Hoc Topic Tracking, which is essentially a combination ofzero-shottopic categorization and the spatio-temporal analysis task. We then propose multiplezero-shotclassification methods to solve this task by building on state-of-the-art language understanding techniques. Next, we picked the best-performing method based on its accuracy on a separate validation dataset and then applied it to a corpus of recent biomedical research articles to track COVID-19 research endeavors across the globe using a spatio-temporal analysis. A demo website has also been developed for HPMs to create custom spatio-temporal visualizations of COVID-19 research trends. The research outcomes demonstrate that the proposedzero-shotclassification methods can potentially facilitate further research on this important subject matter. At the same time, the spatio-temporal visualization tool will greatly assist HPMs and funding agencies in making well-informed policy decisions for advancing scientific research efforts.

Although Artificial Intelligence has become an integral part of modern cybersecurity solutions, data bias and algorithmic bias have made it vulnerable to many cyberattacks, in particular, adversarial attacks where the attacker crafts input of the AI system to exploit the existence of possible bias in the data or algorithms. In this paper, we share our experiences with ongoing work to develop and evaluate a cybersecurity curricular module that demonstrates (a) data bias detection, (b) data bias mitigation, (c) algorithmic bias detection, and (d) algorithmic bias mitigation, using a network intrusion detection problem on real-world data. The module includes lectures and hands-on exercises, using state-of-the-art and open-source bias detection and mitigation software on a real-world dataset. The goal is to identify and mitigate the prevailing conscious/unconscious bias in data and algorithms that the attacker might exploit.

Dengue fever has been considered to be one of the global public health problems of the twenty-first century, especially in tropical and subtropical countries of the global south. The high morbidity and mortality rates of Dengue fever impose a huge economic and health burden for middle and low-income countries. It is so prevalent in such regions that enforcing a granular level of surveillance is quite impossible. Therefore, it is crucial to explore an alternative cost-effective solution that can provide updates of the ongoing situation in a timely manner. In this paper, we explore the scope and potential of a local newspaper-based dengue surveillance system, using well-known data-mining techniques, in Bangladesh from the analysis of the news contents written in the native language. In addition, we explain the working procedure of developing a novel database, using human-in-the-loop technique, for further analysis, and classification of dengue and its intervention-related news. Our classification method has an f-score of 91.45%, and matches the ground truth of reported cases quite closely. Based on the dengue and intervention-related news, we identified the regions where more intervention efforts are needed to reduce the rate of dengue infection. A demo of this project can be accessed at: http://erdos.dsm.fordham.edu:3009/