Daily Archives: July 6, 2026

Authors: Kiran Paul Kanikaram

Abstract: Software testing is among the most time-consuming and challenging stages of the development lifecycle to automate. Effective testing requires contextual awareness, including an understanding of code semantics and the identification of edge cases. The advancement of Large Language Models (LLMs) has enhanced the feasibility of automated testing by enabling unit, end-to-end, and exploratory testing as well as supporting a more agentic Quality Assurance (QA) process. The following paper will discuss the current use of large language models in software testing, with an emphasis on test case creation, bug detection, and self-healing automated systems based on natural language prompts. Although LLMs are providing novel ways to improve testing efficiency, there are certain obstacles that require special attention. They include incorrect output due to hallucinations, incomplete test coverage, and decreased reliability as the software grows. To better understand the obstacles, the example of a checkout module taken from the existing literature is discussed. The results show that while LLM-based testing methods can achieve useful test coverage, they do not necessarily outperform search-based methods. The analysis concludes that the value of LLMs in Quality Assurance is maximized through a human-in-the-loop approach, supported by a five-layer governance framework. As a result, the role of the QA professional is evolving toward that of a test-quality engineer and AI supervisor, requiring an expanded skillset.

DOI: https://doi.org/10.5281/zenodo.21217705

Published by: vikaspatanker

Authors: Research Scholar Mr. Narender Kumar, Associate Professor Dr. Pramod Kumar

Abstract: The primary objective of this research is to develop and evaluate an effective machine learning-based framework for the early detection of ransomware attacks. The study investigates a range of machine learning techniques, including supervised classification, anomaly detection, and clustering methods, to distinguish ransomware activities from legitimate system behavior. It focuses on extracting and analyzing critical behavioral features such as file access patterns, process execution characteristics, and network communication activities to train predictive models capable of achieving high detection accuracy while minimizing false positive rates.

DOI: https://doi.org/10.5281/zenodo.21217459

Published by: vikaspatanker

Authors: Zohaib Ali

Abstract: Image denoising is a foundational stage of the image signal processing (ISP) pipeline: its output quality bounds every downstream task, including demosaicing, compression, retrieval, and recognition. State-of-the-art deep denoisers (DnCNN, FFDNet, CBDNet) achieve strong quality but are typically designed and evaluated for offline, GPU-server settings, leaving embedded and real-time deployment as a secondary concern addressed only through post-hoc compression. This paper designs a denoiser to be lightweight from the outset rather than compressed after the fact. We propose a 3-layer residual convolutional network (5.9K parameters) with an auxiliary noise-level input channel (FFDNet-style conditioning), trained across a range of noise levels rather than a single fixed level. In a controlled study, we first show that a comparable single-noise-trained variant generalizes poorly outside its training noise level (PSNR drops from 28.6 dB at sigma=25 to below classical-filter performance at sigma=50). We then show that noise-level conditioning directly closes this gap: the conditioned model matches or exceeds Gaussian blur and median filtering across sigma in {10, 25, 50} without retraining, using three orders of magnitude fewer parameters than DnCNN. A channel-width ablation (C=16, 24, 32) further shows that quality does not increase monotonically with capacity under a fixed training budget, underscoring that training schedule — not just architecture size — is central to the lightweight-denoising design space. All results are produced by directly executing the accompanying code (no benchmark numbers are copied from other papers); we report exact scope, data, and hardware limitations in Section 6 alongside a concrete roadmap to full-scale benchmark evaluation (BSD68, Set12, SIDD).

Published by: vikaspatanker