The rapid growth of interconnected technologies, especially within the Internet of Things (IoT), has brought about significant security challenges. One crucial aspect of ensuring the security of these systems lies in identifying software vulnerabilities early in the development process. Traditionally, this involved laborious manual code reviews. However, with the rise of powerful AI models like Large Language Models (LLMs), there's hope for automating this critical task. Researchers have been exploring the use of LLMs for Software Vulnerability Detection (SVD), hoping to improve the speed and efficiency of identifying security flaws in code. A key resource in this research is the CVEFixes dataset, a collection of real-world code examples containing both vulnerable and non-vulnerable code snippets. Recent studies have revealed some flaws in this dataset, particularly duplicate entries and inconsistencies, which can hinder the training of accurate LLM models. To address this issue, and to test whether specific transformations can enhance LLMs' ability to detect vulnerabilities, researchers developed SCoPE (Source Code Processing Engine). SCoPE is a tool that normalizes and simplifies C/C++ code, aiming to improve data quality for training LLMs. By processing code through SCoPE, they created a refined version of CVEFixes to evaluate how different data representations affect LLM performance. Researchers then fine-tuned several LLMs on both the original and refined datasets and evaluated their effectiveness in identifying vulnerabilities. The results, while not showing substantial improvements from the code transformations, confirmed the potential of LLMs in vulnerability detection while highlighting the need for better data preprocessing techniques and larger, cleaner datasets. This research underscores the ongoing challenges in applying LLMs to complex software engineering tasks. Future work could focus on improving these models' understanding of code semantics and developing more sophisticated techniques to refine and augment training datasets. Ultimately, more research is needed to explore how LLMs can enhance security best practices within the fast-evolving landscape of modern software development.