Internet of Things (IoT) facilitates the integration between objects and different sensors to provide communication among them without human intervention. However, the extensive demand for IoT and its various applications has continued to grow, coupled with the need to achieve foolproof security requirements. IoT produces a vast amount of data under several constraints such as low processor, power, and memory. These constraints, along with the invaluable data produced by IoT devices, make IoT vulnerable to various security attacks. This paper presents an overview of IoT, its well-known system architecture, enabling technologies, and discusses security challenges and goals. Furthermore, we analyze security vulnerabilities and provide state-of-the-art security taxonomy. The taxonomy of the most relevant and current IoT security attacks is presented for application, network, and physical layers. While most other surveys studied one of the areas of security measures, this study considers and reports on the most advanced security countermeasures within the areas of autonomic, encryption, and learning-based approaches. Additionally, we uncover security challenges that may be met by the research community regarding security implementation in heterogeneous IoT environment. Finally, we provide different visions about possible security solutions and future research directions.

As the range of security attacks increases across diverse network applications, intrusion detection systems are of central interest. Such detection systems are more crucial for the Internet of Things (IoT) due to the voluminous and sensitive data it produces. However, the real-world network produces imbalanced traffic including different and unknown attack types. Due to this imbalanced nature of network traffic, the traditional learning-based detection techniques suffer from lower overall detection performance, higher false-positive rate, and lower minority-class attack detection rates. To address the issue, we propose a novel deep generative-based model called Class-wise Focal Loss Variational AutoEncoder (CFLVAE) which overcomes the data imbalance problem by generating new samples for minority attack classes. Furthermore, we design an effective and cost-sensitive objective function called Class-wise Focal Loss (CFL) to train the traditional Variational AutoEncoder (VAE). The CFL objective function focuses on different minority class samples and scrutinizes high-level feature representation of observed data. This leads the VAE to generate more realistic, diverse, and quality intrusion data to create a well-balanced intrusion dataset. The balanced dataset results in improving the intrusion detection accuracy of learning-based classifiers. Therefore, a Deep Neural Network (DNN) classifier with a unique architecture is then trained using the balanced intrusion dataset to enhance the detection performance. Moreover, we utilize a challenging and highly imbalanced intrusion dataset called NSL-KDD to conduct an extensive experiment with the proposed model. The results demonstrate that the proposed CFLVAE with DNN (CFLVAE-DNN) model obtains promising performance in generating realistic new intrusion data samples and achieves superior intrusion detection performance. Additionally, the proposed CFLVAE-DNN model outperforms several state-of-the-art data generation and traditional intrusion detection methods. Specifically, the CFLVAE-DNN achieves 88.08% overall intrusion detection accuracy and 3.77% false positive rate. More significantly, it obtains the highest low-frequency attack detection rates for U2R (79.25%) and R2L (67.5%) against all the state-of-the-art algorithms.

Managing the System Development Lifecycle (SDLC) is a complex task because of its involvement in coordinating diverse activities, stakeholders, and resources while ensuring project goals are met efficiently. The complex nature of the SDLC process leaves plenty of scope for human error, which impacts the overall business cost. This paper introduces AI-Analyst, an AI-assisted framework developed using the transformer-based model with more than 150 million parameters to assist with SDLC management. It minimizes manual effort errors, optimizes resource allocation, and improves decision-making processes, resulting in substantial cost savings. The statistical analysis shows that it saves around 53.33% of costs in an experimental project. The transformer model has been trained with a uniquely prepared dataset tailored for SDLC through transfer learning. It achieved impressive results, with an accuracy of 91.5%, precision of 91.9%, recall of 91.3%, and an F1-score of 91.5%, demonstrating its high reliability and performance. The perplexity score of 15 further indicates the model’s strong language understanding capabilities to retrieve relations from complex characteristics of Natural Language Processing (NLP). The AI-Analyst framework represents a significant advancement in integrating Large Language Models (LLMs) into SDLC, offering a scalable and cost-effective solution for optimizing business processes.

Strategic cost optimization is a critical challenge for businesses aiming to maintain competitiveness in dynamic markets. This paper introduces Gen-Optimizer, a Generative AI-based framework designed to analyze and optimize business costs through intelligent decision support. The framework employs a transformer-based model with over 140 million parameters, fine-tuned using a diverse dataset of cost-related business scenarios. By leveraging generative capabilities, Gen-Optimizer minimizes inefficiencies, automates cost analysis tasks, and provides actionable insights to decision-makers. The proposed framework achieves exceptional performance metrics, including a prediction accuracy of 93.2%, precision of 93.5%, recall of 93.1%, and an F1-score of 93.3%. The perplexity score of 20.17 demonstrates the model’s superior language understanding and generative abilities. Gen-Optimizer was tested in real-world scenarios, demonstrating its ability to reduce operational costs by 4.11% across key business functions. Furthermore, it aligns with sustainability objectives, promoting resource efficiency and reducing waste. This paper highlights the transformative potential of Generative AI in business cost management, paving the way for scalable, intelligent, and cost-effective solutions.

Environmental sustainability is a global challenge in the face of increasing incidences of disasters affecting communities worldwide. This requires predicting net environmental effects accurately. While various approaches exist, we need more sophisticated prediction models that account for both environmental and social factors. This study presents a proof-of-concept machine learning model for predicting net environmental effects using synthetic data. We developed a multiple linear regression model incorporating nine key features: renewable energy usage, carbon emissions, air quality index, water usage, biodiversity impact, land use, public awareness, and environmental attitudes. We generated a synthetic dataset of 1000 samples using probability distributions and correlation structures derived from environmental literature and expert knowledge. Our model achieved an R-squared value of 0.67, demonstrating moderate predictive power. Feature importance analysis revealed renewable energy usage (coefficient = 0.71) and public awareness (coefficient = 0.44) as significant positive factors influencing environmental outcomes. Model validation included residual analysis and feature importance assessment, with results suggesting reasonable performance within linear regression constraints. Limitations of our study include reliance on synthetic data, assumption of linear relationships between variables, and limited environmental factors. Notwithstanding, our findings provide insights for environmental policymaking, particularly regarding renewable energy adoption and public awareness campaigns. Future work could focus on incorporating real-world data, exploring non-linear modeling approaches, and expanding the feature set to capture more complex environmental interactions. Our research contributes to data-driven environmental assessment by demonstrating the feasibility of combining both physical and social factors in predictive modeling.

Insider threats pose a significant and growing risk to organizational cybersecurity, with recent studies indicating a 47% increase in insider incidents from 2018 to 2022. This paper presents a comparative analysis of unsupervised and supervised machine learning approaches for detecting potential insider threats through network traffic anomaly identification. We develop and evaluate an Isolation Forest (unsupervised) and a Random Forest (supervised) model, training them on a simulated dataset representing six months of network logs from a mid-sized company. Our study introduces a unique feature set combining traditional network metrics with temporal and behavioral indicators, enhancing the models' detection capabilities. Results show that the Random Forest classifier outperforms the Isolation Forest, with F1-scores of 0.6425 and 0.4624, respectively. However, the unsupervised approach shows promise in scenarios lacking labeled data. Key findings reveal that increased connection frequency and data transfer volume are critical indicators of potential threats, with temporal patterns also playing a significant role. This study provides valuable insights into the strengths and limitations of each approach, offering practical implications for real-world digital forensics investigations. We contribute to the field by proposing a hybrid approach that leverages the strengths of both methods, potentially improving the accuracy and adaptability of insider threat detection systems. These findings pave the way for more robust, context-aware cybersecurity measures in the digital age.