Abstract:Machine learning (ML) is a rapidly developing area of medicine that uses significant resources to apply computer science and statistics to medical issues. ML's proponents laud its capacity to handle vast, complicated, and erratic medical data. It's common knowledge that attackers might cause misclassification by deliberately creating inputs for machine learning classifiers. Research on adversarial examples has been extensively conducted in the field of computer vision applications. Healthcare systems are thought to be highly difficult because of the security and life-or-death considerations they include, and performance accuracy is very important. Recent arguments have suggested that adversarial attacks could be made against medical image analysis (MedIA) technologies because of the accompanying technology infrastructure and powerful financial incentives. Since the diagnosis will be the basis for important decisions, it is essential to assess how strong medical DNN tasks are against adversarial attacks. Simple adversarial attacks have been taken into account in several earlier studies. However, DNNs are susceptible to more risky and realistic attacks. The present paper covers recent proposed adversarial attack strategies against DNNs for medical imaging as well as countermeasures. In this study, we review current techniques for adversarial imaging attacks, detections. It also encompasses various facets of these techniques and offers suggestions for the robustness of neural networks to be improved in the future.

What problem does this paper attempt to address?

### What problems does this paper attempt to solve? This paper aims to explore and analyze adversarial attacks against deep neural networks (DNNs) in medical image analysis (MedIA) and evaluate the defense capabilities of existing techniques against these attacks. Specifically, the goals of the paper include: 1. **Identifying vulnerabilities**: Provide an in - depth overview covering various adversarial attack strategies against DNNs in medical image analysis to identify potential vulnerabilities. 2. **Evaluating model robustness**: Examine the application of existing DNN models in medical image analysis and their resistance to adversarial threats. 3. **Testing model performance**: Evaluate the effectiveness of these DNN models when facing adversarial attacks. 4. **Proposing improvement suggestions**: Provide improvement suggestions for enhancing the security of medical image analysis systems to ensure their reliability and accuracy in critical areas. #### Background and importance In recent years, machine learning (ML), especially deep learning (DL), has made significant progress in the field of medical image analysis. However, as these technologies are applied more and more widely, security and reliability issues have become increasingly prominent. Adversarial attacks deceive machine - learning models by deliberately tampering with input data, which may lead to misclassification or misdiagnosis. This type of attack has been widely studied in the field of computer vision, but relatively less explored in the field of medical image analysis. Since medical image analysis is directly related to the life and health of patients, the security and accuracy of its models are crucial. #### Research methods To achieve the above - mentioned goals, the paper adopts the following methods: - **Literature review**: Reviewed existing adversarial attack and defense techniques, with particular attention to techniques developed within the last two years. - **Experimental verification**: Conducted experiments using four publicly available benchmark datasets (such as Messidor1, ISIC, ChestX - ray 14, and the COVID - 19 database) to verify the effects of different attack and defense strategies. - **Classification system**: Proposed a new classification system that classifies adversarial defense strategies into white - box attacks, semi - white - box attacks, and black - box attacks according to the application scenario. #### Results and discussion Through experiments, the authors demonstrated the following: 1. **Effectiveness of the SSAT module**: The SSAT module can significantly improve the model's adversarial robustness without reducing the classification accuracy of clean images. 2. **Impact of adversarial samples**: The generated adversarial samples can effectively deceive DNN models, resulting in incorrect diagnosis results. 3. **Comparison of defense strategies**: Different defense strategies perform differently in adversarial attacks, and some strategies are more effective under specific conditions. In conclusion, through comprehensive analysis and experimental verification, this paper reveals the vulnerabilities of DNNs in medical image analysis and provides important references and improvement suggestions for future research.

Securing the Diagnosis of Medical Imaging: An In-depth Analysis of AI-Resistant Attacks

Adversarial attacks and defenses on AI in medical imaging informatics: A survey

Adversarial examples: attacks and defences on medical deep learning systems

Robust Medical Diagnosis: A Novel Two-Phase Deep Learning Framework for Adversarial Proof Disease Detection in Radiology Images

Securing Medical Imaging: A Comprehensive Review and Prospective Paths for Image Authentication

How Resilient Are Deep Learning Models in Medical Image Analysis? The Case of the Moment-Based Adversarial Attack (Mb-AdA)

Defending Against Adversarial Attacks On Medical Imaging Ai System, Classification Or Detection?

The Impact of Simultaneous Adversarial Attacks on Robustness of Medical Image Analysis

Understanding adversarial attacks on deep learning based medical image analysis systems

DFT-Based Adversarial Attack Detection in MRI Brain Imaging: Enhancing Diagnostic Accuracy in Alzheimer's Case Studies

Adversarial Robust Image Processing in Medical Digital Twin

Survey on Adversarial Attack and Defense for Medical Image Analysis: Methods and Challenges

Comparing Detection Schemes for Adversarial Images against Deep Learning Models for Cancer Imaging

Practical Fast Gradient Sign Attack against Mammographic Image Classifier

Adversarial Attacks Against Medical Deep Learning Systems

Adversarial attack vulnerability of medical image analysis systems: Unexplored factors

Frequency Constraint-Based Adversarial Attack on Deep Neural Networks for Medical Image Classification.

Adversarial Attacks and Defences for Skin Cancer Classification

Robust Convolutional Neural Networks Against Adversarial Attacks on Medical Images

Adversarial Neural Networks in Medical Imaging Advancements and Challenges in Semantic Segmentation

Adversarial Deep Learning: A Survey on Adversarial Attacks and Defense Mechanisms on Image Classification