The last decade has seen a rise of Deep Learning with its applications ranging across diverse domains. But usually, the datasets used to drive these systems contain data which is highly confidential and sensitive. Though, Deep Learning models can be stolen, or reverse engineered, confidential training data can be inferred, and other privacy and security concerns have been identified. Therefore, these systems are highly prone to security attacks. This study highlights academic research that highlights the several types of security attacks and provides a comprehensive overview of the most widely used privacy-preserving solutions. This relevant systematic evaluation also illuminates potential future possibilities for study, instruction, and usage in the fields of privacy and deep learning.
翻译:在过去的十年中,深层学习的兴起,其应用范围遍及不同领域,但通常用于驱动这些系统的数据集包含高度机密和敏感的数据。虽然深层学习模式可以被盗,或者反向设计,但可以推断出机密培训数据,并查明其他隐私和安全关切。因此,这些系统极易受到安全攻击。本研究报告强调了突出几种类型的安全攻击的学术研究,并全面概述了最广泛使用的隐私保护解决方案。这一相关的系统评估还揭示了今后在隐私和深层学习领域研究、指导和使用的可能性。