International Journal of Intelligent Systems and Data Science

Vol. 1 No. 3

2026
June 2026 · 15 articles · Volume 1, Issue 3
Back to archives
ISSUE OVERVIEW

Intelligent systems. Real-world impact.

This issue presents research at the intersection of artificial intelligence, data science, and modern computational systems. The contributions explore advances in large and small language models, distributed cognitive systems, blockchain technologies, cybersecurity, container orchestration, embedded systems, dynamic optimization, and machine learning applications in agriculture and audio synthesis. Combining theoretical innovations with practical implementations, the articles highlight the growing convergence of intelligent algorithms and scalable computing infrastructures, reinforcing the importance of interdisciplinary research in addressing complex scientific and engineering challenges.

Published

June 2026

Contributions

15 articles

Access

Open access

June 2026 · 15 articles
Issue Cover
EXPLORE THIS ISSUE
Table of contents
15 articles

Explore the research published in this issue of ijisds.

IN THIS ISSUE
Kernel-Safe Data Probes: Context-Aware Address Tagging with On-Demand Translation for OS Modules

Monisha Rengaraj

This paper presents a software-only data-access monitoring architecture for operating-system kernels that attaches context to memory references and activates selective instrumentation through on-demand binary translation. The design encodes per-object metadata into pointer space to enable type- and field-aware checks at instruction granularity while keeping common kernel paths free of persistent overhead. A dual execution path resolves watched references either via fast inline masking or via trap-driven translation when events are rare, balancing precision and cost under real kernel workloads. The implementation for x86-64 Linux modules supports practical tools including overflow and bounds checking, read-before-write and lifetime misuse detection, leak discovery with selective shadow state, and field-level policy enforcement for tamper-resistant subsystems. Evaluation with microbenchmarks and storage-stack file operations quantifies baseline translation cost and the incremental overheads of precise bounds and fault-mediated paths, explaining how selective activation and object scoping reduce exception storms in hot kernel code. Results show that the approach yields actionable diagnostics for module reliability and security while preserving predictable performance through configurable granularity and targeted instrumentation.

PDF
Multicore SMT-BMC for Embedded Co-Design: MaxSMT-Aided Hardware/Software Partitioning at Scale

Monisha Rengaraj

This paper targets hardware/software partitioning in embedded systems as a first-class design automation problem, presenting a multicore Bounded Model Checking (BMC) workflow that accelerates exact optimization under real system constraints. The proposed approach parallelizes Satisfiability Modulo Theories (SMT)-based verification with Open Multi-Processing (OpenMP) and augments it with a Maximum Satisfiability Modulo Theories (MaxSMT) backend (vZ) to minimize hardware cost while satisfying software and communication constraints on task-graph models used in co-design. A family of implementations, including the sequential Efficient SMT-Based Context-Bounded Model Checker (ESBMC), multicore ESBMC with sequential, worker, and binary-search controllers, and an ESBMC-(vZ) integration, is developed to exploit contemporary multicore processors without modifying the underlying solver backends. Comparative evaluation against Integer Linear Programming (ILP) and Genetic Algorithm (GA) approaches across standard embedded-system benchmarks (e.g., CRC32, Patricia, Dijkstra, Clustering, RC6, Fuzzy, and Mars) quantifies runtime, optimality, and scalability. The results demonstrate that the MaxSMT-based ESBMC approach achieves superior performance on small-to-medium problem instances, whereas the multicore ESBMC with a parallel binary-search controller (ESBMC-PB) provides substantial speedups for larger task graphs. These findings offer practical guidance for hardware/software co-design workflows: employ ESBMC-(vZ) when exact optimality and rapid turnaround are priorities, utilize multicore ESBMC-PB for large design spaces, and reserve GA for scenarios where approximate solutions are acceptable. Overall, the proposed contributions establish a practical Computer Engineering (CE) design-automation pipeline that effectively integrates formal optimization with parallel verification to address hardware/software trade-offs in embedded system architectures.

PDF
Artificial Intelligence: Adaptive Certainty Propagation in Distributed Cognitive Systems

Shiva Bommakanti

Distributed artificial intelligence (AI) systems require protocols with scalability and fault tolerance for making decisions asynchronously over networks. Through the adaptive propagation of a certain strength and the recursive structure of consensus summits, a new consensus protocol is proposed in this paper, which is robust and flexible. The proposed framework allows agents to reach agreement with various confidence levels rather than using binary commitments like in traditional approaches. Moreover, these confidence levels are based on validator reliability and consistency. By utilizing a Directed Acyclic Graph (DAG), the flow of information can be organized effectively. This allows for increased scaling and a reduction in the costs of communication. Furthermore, aggregate summit proof validation can be done efficiently as the networks get enhanced. With iteration of the recursive summit mechanism, more trust gets established in the system's decision, and it becomes robust to noisy observations, network latencies, and malicious behavior among parties. The protocol performs its task asynchronously and does not require the existence of a global clock. Moreover, the protocol achieves Byzantine fault tolerance in the presence of up to one-third malicious validators. The adaptive confidence model enhances collective decision-making across heterogeneous AI ecosystems, making the framework suitable for decentralized machine learning platforms and large language model deployments. The method achieves the scalable verification of finality for strong consensus. These attributes make the protocol a potential solution for next-generation distributed artificial intelligence systems.

PDF
CRD-Shim: Declarative Container Runtime Orchestration via Kubernetes CRDs for Docker & containerd

Jitendra Gupta

Kubernetes is now an essential platform for orchestrating distributed containers based on a declarative API approach for deploying and managing containerized distributed applications. Conventionally, Kubernetes orchestration goes through intermediary layers like Pods, kubelet, and CRI, incurring extra latency, overhead, and complexity. This paper introduces CRD-Shim, a Kubernetes-native interface to directly orchestrate container runtimes like Docker and containerd declaratively using CRDs. CRD-Shim orchestrates runtimes by transforming Kubernetes CRDs into calls to their underlying native API calls without involving Pods, kubelet, and CRI. The architecture is compatible with container-specific resource objects and also provides high-level orchestration functionalities such as a controller for resource replication (ReplicaSet equivalent) purely through runtimes. Our experiment results show that CRD-Shim reduces the container startup and update latency to nearly 62% and 30%, respectively, compared with a single-node Kubernetes deployment, and decreases its runtime memory usage to almost 66% of that. CRD-Shim can serve as an effective option for scenarios where direct runtime management and low-latency orchestration are more critical than general container ecosystem management like edge computing, IoT and single-node systems.

PDF
Pi-Scale Mediation: An Empirical Study on Tangible Container Clusters as Pedagogical Catalysts in Distributed Systems Education

Sheshukumar Vangala

This paper presents an empirical investigation into the efficacy of tangible, small-scale cloud computing clusters as pedagogical mediators for bridging the conceptual gap in distributed systems education. While cloud computing and microservices have reshaped the software industry, academic curricula often lag due to the abstract nature and high cost of real-world infrastructure. We designed, built, and evaluated KubeCloud, a low-cost, portable cluster comprising four Raspberry Pi nodes running Docker and Kubernetes. This physical learning object served as the centrepiece of a seven-week, problem-based learning activity for engineering students. Through iterative experimentation including cable-pulling fault injection, application-level resilience testing with circuit breakers, and infrastructure-level replication studies, we quantified the cluster's impact on student comprehension. Evaluations based on SOLO and Bloom's taxonomies demonstrated a measurable progression from pre-structural to relational understanding. Findings indicate that tangible clusters significantly enhance the visibility of cloud abstractions, improve grasp of resilience concepts, and align engineering students' sensory-visual learning preferences with complex distributed systems paradigms. The study concludes that such mediated, hands-on environments are not only viable but essential for preparing graduates with the practical, failure-embracing mindset required by modern IT landscapes.

PDF
Latent Phrase-Aware Generative Modeling for Expressive Symbolic Audio Synthesis

Apeksha Bhuekar

Constructing expressive symbolic music is a hard task. A good generator must take into account long-range musical structure and fine-grained performance features at the same time. Traditional sequence-based methods typically focus on pitch and timing information while providing limited support for expressive techniques such as bends, slides, vibrato and dynamic articulation. In this paper, we propose a novel generative framework that employs a compact tokenization scheme and phrase-aware latent alignment mechanism to enhance the quality and controllability of symbolic audio synthesis. The tokenization scheme efficiently represents both basic musical events and expressive performance attributes with a limited vocabulary, resulting in substantial computational savings without semantic loss. The phrase-level latent representations are injected into the transformer attention through a KL-divergence-based bias, such that variable-length musical phrases' structural dependencies can be learned. By applying sequence regularization and a repetition-aware loss, a multi-objective optimization framework enhances generation quality by minimizing redundant expressive patterns. Through experimental evaluation on a guitar tablature dataset, we show that our model surpasses established transformer-based baselines on a number of aspects: perplexity, diversity, speed, and expressiveness. These findings prove the proposed framework’s efficiency in generating coherence, expressiveness and computational efficiency in symbolic music.

PDF
End‑to‑End Data‑Mining Pipelines with Fine‑Tuned Language Models for Reliable Text Classification

Manasi Bendale

This paper proposes a practical data-mining workflow for supervised text classification that combines pretrained neural architectures with a systematic and reproducible development process that covers the complete model life-cycle from corpus setup and data pre-processing to training, evaluation and interpretation. The proposed workflow minimizes laborious manual feature extraction by leveraging the representation power of the pretrained models while adhering to established protocols for data partitioning, model selection and performance measurement. We used the standard metrics for evaluation, such as precision, recall, F1-score, confusion matrix, and learning-curve analysis, to characterize classification performance, assess generalization performance and study the impact of training data size on predictive accuracy. Our experimental results using different datasets show consistent improvements over standard machine learning baselines while remaining computationally efficient and reproducible. It also provides practical advice for GPU-accelerated training to lower execution time and improve scaling to larger collections of texts. In order to facilitate transparent model evaluation, our workflow incorporates local model-agnostic explanation techniques that enable us to clarify which input tokens are responsible for a prediction outcome, thus allowing for detailed error analysis and validation of a given classification decision. Taking everything into account, the offered framework provides a portable, efficient and interpretable toolbox for accurate text classification design in a common statistical computing environment.

PDF
Integrating Trusted Types with Solid.js and Cypress: Enhancing Web Application Security

Santosh Vemula

This paper studies Trusted Types, which is an updated browser restriction to mitigate Cross-Site Scripting (XSS) attacks for the Solid.js framework and the Cypress e2e testing environment. The research commences with the migration of the Solid.js build system from the Rollup bundler over to Vite, requiring minimal changes to the existing codebase while enhancing the efficiency of the development workflow. In order to achieve full Trusted Types compatibility, we relied on custom Vite and Solid.js versions. Also, we made necessary adjustments, such as renaming JavaScript source files from .js to .jsx and implementing Trusted Types policies for generating third-party API content. Furthermore, the study describes a custom Cypress plugin used to validate the enforcement of Trusted Types in automated testing. A developer can easily configure Content Security Policy (CSP) headers and detect Trusted Types violations in real-time with the help of the plugin. It aids in making security testing more dependable. As shown in the experiments, the proposed integration strengthens application resilience against client-side injection attacks with the added benefit of efficient development and flexible testing. While usage of Trusted Types in the open source ecosystem is not widespread, it shows great potential for enhancing the security of modern web applications. The future work will extend Trusted Types support to large-scale real-world applications, improve compatibility with frameworks, and promote wide-scale adoption of secure-by-default web development practices.

PDF
Perturbation-Driven Visual Analytics for Probing InternalRepresentations in Neural Language Models

Eeshwar Pasula

This paper introduces an interactive visual analytics framework designed to support exploratory analysis of neural network models for natural language inference through a perturbation-driven paradigm. Rather than treating trained models as static black boxes, the system enables users to dynamically manipulate inputs, internal attention mechanisms, and output predictions while observing corresponding changes across the processing pipeline. The interface integrates multiple coordinated views including bipartite graph and matrix representations of attention, barycentric coordinate plots for probabilistic predictions, and a pipeline visualization for tracking parameter updates to support hypothesis formation and causal reasoning about model behavior. A constrained optimization procedure, inspired by the margin-infused relaxed algorithm, allows users to correct erroneous predictions while minimizing parameter deviation and to compare the relative influence of encoder, attention, and classifier stages on prediction outcomes. The system also overlays syntactic dependency structures onto attention visualizations, enabling grammar-guided simplification and facilitating investigation of the relationship between linguistic structure and learned alignments. Evaluation with NLP researchers demonstrates that the tool supports a range of analytical tasks, including stability assessment, error diagnosis, attention editing, and comparative analysis of model components. The framework is implemented as a lightweight Python library that integrates with existing PyTorch models, lowering the barrier to adoption for routine model interrogation.

PDF
Enabling Decentralized, Programmable Order Flow Auctions for MEV Mitigation in DeFi

Muneebuddin Syed

A decentralized auction system that is decentralized is proposed to solve the many limitations that occur with traditional DeFi order flow management. Those limitations often come from centralized coordination and poor transaction execution. The framework enables the development of intent-driven auctions to improve liquidity discovery while allowing independent solvers to join seamlessly, separating the execution logic from application-specific functionality. As a result, one does not have to depend on intermediaries for routing or executing the transaction. The architecture is also compatible with alternative auction mechanisms. This not only permits a broad range of customizable decentralized applications (dApps) but also provides a path to incorporate future developments, such as secure computation and cryptographic advances, ensuring the continuous evolution of ecosystems. Its modular design fosters interoperability between applications, allowing different parts to evolve independently but still work well together in the larger ecosystem. Simultaneously, it encourages transparency in competition among solvers, strengthens decentralization across the MEV supply chain, and mitigates opportunities for rent extraction by centralised actors. Faster execution of transactions and increased coordination of liquidity will reduce costs and improve efficiency across blockchain networks. Moreover, the architecture provides a flexible and workable mechanism to manage cross-chain order flows. It enables safe, reliable and efficient execution, while respecting the foundational decentralized principles of modern blockchain systems.

PDF
Showing 1 to 10 of 15 articles
Research, freely accessible

Read the articles online or download individual PDFs from this issue.

Vol. 1 No. 3 (2026) | International Journal of Intelligent Systems and Data Science