Recursive Decay in AI-Generated Code: An Empirical Study of Security Logic Degradation Through Iterative LLM Refactoring

Kaustuv Sharma; Yatu Rani; Naveen Yadav

Recursive Decay in AI-Generated Code: An Empirical Study of Security Logic Degradation Through Iterative LLM Refactoring

Author(s)	Mr. Kaustuv Sharma, Yatu Rani, Naveen Yadav
Country	India
Abstract	Large language models (LLMs) are increasinglyintegrated into software development workflows for code gen-eration, refactoring, and optimization. Although the productivitybenefits of such tools are well-documented, their long-termimpact on code correctness and security under iterative usehas received limited empirical attention. This paper introducesthe Recursive Code Decay Experiment, an empirical frameworkthat quantifies how security-critical logic erodes when code isrepeatedly processed by generative AI models under variousrefactoring objectives.We evaluate LLM-driven refactoring across seven models—Llama 3.3 70B, Llama 3.1 8B, GPT-OSS 120B, GPT-OSS20B, Gemma 3 4B, Qwen 3 8B, and DeepSeek-R1 8B—usingaggressive, neutral, and preservation-aware prompting strate-gies. Security integrity is audited at each generation usingthe Guardrail Pattern Verification (GPV) algorithm, an AST- based structural analyzer capable of detecting five categories ofdefensive programming patterns. Our findings reveal three dis-inct decay archetypes: monotonic, oscillatory, and paradoxical.Notably, prompt formulation outweighs model size as a predictorof decay resistance. DeepSeek-R1 uniquely exhibited completesecurity collapse even under an explicit preservation directive, highlighting a critical gap between semantic understandingandstructural preservation. These results underscore the necessityof automated structural auditing in AI-assisted developmentpipelines.
Keywords	AI Code Generation, LLM Refactoring, Code Security, AST Analysis, Prompt Engineering, Guardrail Pattern Verification, Recursive Decay, Technical Debt
Field	Computer > Artificial Intelligence / Simulation / Virtual Reality
Published In	Volume 8, Issue 3, May-June 2026
Published On	2026-05-13

View / Download PDF File

E-ISSN 2582-2160

doi

CrossRef DOI is assigned to each research paper published in our journal.

IJFMR DOI prefix is
10.36948/ijfmr

Downloads

Research Paper Format Copyright Permission Form and Undertaking Form Cover Page Vol 8 Isu 2 Cover Page Vol 8 Isu 1 Cover Page Vol 7 Isu 6

All research papers published on this website are licensed under Creative Commons Attribution-ShareAlike 4.0 International License, and all rights belong to their respective authors/researchers.

CC-BY-SA

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Recursive Decay in AI-Generated Code: An Empirical Study of Security Logic Degradation Through Iterative LLM Refactoring

Share this