[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117169-en":3,"doc-seo-117169-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},117169,137441390410,"Hazel","https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984",8,"Research & Report","Python Fuzzing for Trustworthy Machine Learning Frameworks","Ensuring the security and reliability of machine learning frameworks is crucial for building trustworthy AI-based systems. The paper presents a dynamic analysis pipeline for Python projects using the Sydr-Fuzz toolset, combining fuzzing, corpus minimization, crash triaging, and coverage collection. It emphasizes crash triage and severity estimation to prioritize the most critical vulnerabilities and integrates the pipeline into GitLab CI. Attack surfaces are analyzed to build fuzz targets for PyTorch, TensorFlow, and related projects such as h5py. The approach discovers three new bugs and proposes fixes.","arXiv :2403 . 12723v2 [ cs .CR] 23 Dec 2024  \nPython Fuzzing for Trustworthy Machine Learning Frameworks  \nIlya Yegorov 1 ,2[0000􀀀0003􀀀2158􀀀0414], Eli Kobrin1 ,2[0000􀀀0002􀀀6035􀀀0577], Darya Parygina 1 ,2[0000􀀀0002􀀀4029􀀀0853], Alexey Vishnyakov 1[0000􀀀0003􀀀1819􀀀220X], and Andrey Fedotov 1[0000􀀀0002􀀀8838􀀀471X]  \n1 Ivannikov Institute for System Programming of the RAS  \n2 Lomonosov Moscow State University  \n{Yegorov_Ilya,kobrineli,pa_darochek,vishnya,[fedotoff}@ispras.ru](fedotoff}@ispras.ru)  \n[Abstract.](Abstract. Ensuring the security and reliability of machine learning)[ Ensuring the security and reliability of machine learning](Abstract. Ensuring the security and reliability of machine learning)[ ](Abstract. Ensuring the security and reliability of machine learning)[frameworks is crucial for building trustworthy AI-based systems. Fuzzing](frameworks is crucial for building trustworthy AI-based systems. Fuzzing), a popular technique in secure software development lifecycle (SSDLC), can be used to develop secure and robust software. Popular machine learning frameworks such as PyTorch and TensorFlow are complex and written in multiple programming languages including C/C++ and Python.  \nWe propose a dynamic analysis pipeline for Python projects using the Sydr-Fuzz toolset. Our pipeline includes fuzzing, corpus minimization, crash triaging, and coverage collection. Crash triaging and severity estimation are important steps to ensure that the most critical vulnerabilities are addressed promptly. Furthermore, the proposed pipeline is integrated in GitLab CI. To identify the most vulnerable parts of the machine learning frameworks, we analyze their potential attack surfacesand develop fuzz targets for PyTorch, TensorFlow, and related projects such as h5py. Applying our dynamic analysis pipeline to these targets, we were able to discover 3 new bugs and propose 􀀜xes for them.  \nKeywords: Fuzzing · Trustworthy AI · Machine learning framework · TensorFlow · PyTorch · Python · Arti􀀜cial intelligence · Crash triage  \n· Dynamic analysis · Secure software development lifecycle · SSDLC · Computer security  \n1 Introduction  \nArti􀀜cial Intelligence (AI) has been the focus of signi􀀜cant attention in recent years due to its transformative potential across a wide range of industries. However, with the increasing prevalence of AI-based systems, security concerns have also emerged. Ensuring the reliability and safety of these systems is paramount, and the 􀀜eld of secure AI has emerged as a multi-disciplinary area of computer science dedicated to achieving this goal.  \nAI-based systems are typically built on top of machine learning (ML) frameworks, such as TensorFlow [23] and PyTorch [19], which provide the necessary  \nYegorov I., Kobrin E., Parygina D., Vishnyakov A., Fedotov A. Python fuzzing for trustworthy machine learning frameworks. Zapiski Nauchnykh Seminarov POMI, St. Petersburg Branch of the V.A. Steklov Mathematical Institute of the Russian Academy of Sciences, Vol. 530, 2023, pp. 38􀀕50 .  \nJournal of Mathematical Sciences, 2024 Springer Nature Switzerland AG, Vol. 285, No.  \n2, October, 2024 . DOI 10.1007/s10958-024-07424-2 .  \n2 I. Yegorov et al.  \ntools to implement and train machine learning models. The growing complexity and pervasiveness of these frameworks have made them a prime target for attackers seeking to exploit vulnerabilities for malicious purposes. The potential consequences of a successful attack on an AI-based system are signi􀀜cant, ranging from compromising the integrity and con􀀜dentiality of sensitive data to causing physical harm or 􀀜nancial loss.  \nDeveloping secure and trustworthy ML frameworks is a challenging task, given the large and complex codebase of popular ML frameworks, which are developed in multiple programming languages such as C/C++ and Python. Secure software development life cycle (SSDLC) is commonly used to provide safety and code quality for open source and commercial projects. Dynamic analysis, such","cbCairfAUtctgce1","https://ap.wps.com/l/cbCairfAUtctgce1","pdf",174345,1,12,"English","en",105,"# Introduction\n## Related Work\n## Atheris\n# Dynamic Analysis Pipeline\n## Attack Surface Analysis\n# Fuzzing Results\n# Conclusion","[{\"question\":\"What is the main goal of the proposed work?\",\"answer\":\"The work targets security and reliability of machine learning frameworks by using fuzzing-based dynamic analysis to find vulnerabilities in complex framework codebases.\"},{\"question\":\"What steps are included in the Sydr-Fuzz dynamic analysis pipeline?\",\"answer\":\"The pipeline includes fuzzing, corpus minimization, crash triaging, and coverage collection, with crash triage used for severity estimation and prioritization.\"},{\"question\":\"Which machine learning frameworks and projects are evaluated with fuzzing targets?\",\"answer\":\"The paper develops fuzz targets for PyTorch, TensorFlow, and related projects such as h5py, guided by analysis of their potential attack surfaces.\"}]","Python Fuzzing for Trustworthy Machine Learning Frameworks | PDF",1785674198,30,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"python-fuzzing-for-trustworthy-machine-learning-frameworks","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/python-fuzzing-for-trustworthy-machine-learning-frameworks/117169/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the main goal of the proposed work?","Question",{"text":75,"@type":76},"The work targets security and reliability of machine learning frameworks by using fuzzing-based dynamic analysis to find vulnerabilities in complex framework codebases.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What steps are included in the Sydr-Fuzz dynamic analysis pipeline?",{"text":80,"@type":76},"The pipeline includes fuzzing, corpus minimization, crash triaging, and coverage collection, with crash triage used for severity estimation and prioritization.",{"name":82,"@type":73,"acceptedAnswer":83},"Which machine learning frameworks and projects are evaluated with fuzzing targets?",{"text":84,"@type":76},"The paper develops fuzz targets for PyTorch, TensorFlow, and related projects such as h5py, guided by analysis of their potential attack surfaces.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":29,"slug":121},"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]