Enhancing Biosecurity in Tamper-Resistant Large Language Models With Quantum Gradient Descent

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hai, Fahmida, Nirzhor, Saif, Khan, Rubayat, Roosan, Don
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916809076113408
author Hai, Fahmida
Nirzhor, Saif
Khan, Rubayat
Roosan, Don
author_facet Hai, Fahmida
Nirzhor, Saif
Khan, Rubayat
Roosan, Don
contents This paper introduces a tamper-resistant framework for large language models (LLMs) in medical applications, utilizing quantum gradient descent (QGD) to detect malicious parameter modifications in real time. Integrated into a LLaMA-based model, QGD monitors weight amplitude distributions, identifying adversarial fine-tuning anomalies. Tests on the MIMIC and eICU datasets show minimal performance impact (accuracy: 89.1 to 88.3 on MIMIC) while robustly detecting tampering. PubMedQA evaluations confirm preserved biomedical question-answering capabilities. Compared to baselines like selective unlearning and cryptographic fingerprinting, QGD offers superior sensitivity to subtle weight changes. This quantum-inspired approach ensures secure, reliable medical AI, extensible to other high-stakes domains.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19086
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Enhancing Biosecurity in Tamper-Resistant Large Language Models With Quantum Gradient Descent
Hai, Fahmida
Nirzhor, Saif
Khan, Rubayat
Roosan, Don
Molecular Networks
This paper introduces a tamper-resistant framework for large language models (LLMs) in medical applications, utilizing quantum gradient descent (QGD) to detect malicious parameter modifications in real time. Integrated into a LLaMA-based model, QGD monitors weight amplitude distributions, identifying adversarial fine-tuning anomalies. Tests on the MIMIC and eICU datasets show minimal performance impact (accuracy: 89.1 to 88.3 on MIMIC) while robustly detecting tampering. PubMedQA evaluations confirm preserved biomedical question-answering capabilities. Compared to baselines like selective unlearning and cryptographic fingerprinting, QGD offers superior sensitivity to subtle weight changes. This quantum-inspired approach ensures secure, reliable medical AI, extensible to other high-stakes domains.
title Enhancing Biosecurity in Tamper-Resistant Large Language Models With Quantum Gradient Descent
topic Molecular Networks
url https://arxiv.org/abs/2506.19086