analiza aiqbench
This survey comprehensively maps the landscape of model inversion attacks, analyzing how adversaries reconstruct private training data from trained neural networks across images, text, and graphs. It critically evaluates existing defenses by comparing threat models, attacker assumptions, and the inherent trade-offs between privacy preservation and model utility. The work is essential for researchers developing robust machine learning systems that can withstand sophisticated inference threats without sacrificing performance.
Model Inversion Attacks: A Survey of Approaches and Countermeasures
arXiv:2411.10023v3 Announce Type: replace Abstract: Deep neural networks have enabled numerous studies and applications on both Euclidean data, such as images and text, and non-Euclidean data, such as graphs. Because these networks may process private data, their deployment raises concerns about privacy leakage. Model inversion attacks (MIAs) exploit access to a trained model to reconstruct training examples or infer privacy-sensitive characteristics represented by the model. The effectiveness of MIAs has been demonstrated in various domains, including images, text, and graphs. These attacks highlight the vulnerability of neural networks and raise awareness about the risk of privacy leakage within the research community. This survey provides a threat-model- and assumption-aware synthesis of attacks and defenses. We compare exposed interfaces, attacker knowledge, reconstruction priors and targets, failure modes, deployment constraints, and privacy-utility trade-offs, while highlighting modeling principles, optimization challenges, and future directions. We also maintain an evolving repository of relevant research at https://github.com/AndrewZhou924/Awesome-model-inversion-attack.
Abstract:Deep neural networks have enabled numerous studies and applications on both Euclidean data, such as images and text, and non-Euclidean data, such as graphs. Because these networks may process private data, their deployment raises concerns about privacy leakage. Model inversion attacks (MIAs) exploit access to a trained model to reconstruct training examples or infer privacy-sensitive characteristics represented by the model. The effectiveness of MIAs has been demonstrated in various domains, including images, text, and graphs. These attacks highlight the vulnerability of neural networks and raise awareness about the risk of privacy leakage within the research community. This survey provides a threat-model- and assumption-aware synthesis of attacks and defenses. We compare exposed interfaces, attacker knowledge, reconstruction priors and targets, failure modes, deployment constraints, and privacy-utility trade-offs, while highlighting modeling principles, optimization challenges, and future directions. We also maintain an evolving repository of relevant research at this https URL.
Machine Learning (cs.LG)
arXiv:2411.10023 [cs.LG]
(or arXiv:2411.10023v3 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2411.10023 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 239 cuvinte
Subiecte conexe
Research Note: Filtering Subversion-Relevant Information From Pretraining Data Is Feasible
LessWrong · 5 oct. 2026
Implications of Continual Learning for LLM Agents: Introduction
LessWrong · 12 iun. 2026
Lattice-Based Cryptography Explained
Quantum Zeitgeist · 15 aug. 2026
State of Pandemic Early Warning
LessWrong · 24 sept. 2026