隐形标记:通过不确定性引导的后门对医疗细分进行无害和隐蔽的所有权验证
概括
保护AI医疗细分模型至关重要. StealthMark提供了一种新,无害的方法,通过嵌入QR码水印在模型说明中来验证所有权,确保性能和知识产权安全.
科学领域:
- 人工智能的人工智能
- 医疗成像医学成像
- 计算机视觉 计算机视觉
背景情况:
- 对人工智能培训的医疗数据注释是昂贵的,并且受到专家短缺和隐私问题的限制.
- 训练有素的医疗细分模型是有价值的知识产权,但缺乏强大的保护方法.
- 现有的模型保护技术在很大程度上忽视了对医疗图像分析至关重要的细分模型.
研究的目的:
- 提出StealthMark,一种新的,隐蔽的,无害的方法来验证医疗细分模型的所有权.
- 为了实现黑子所有权验证,而不会影响模型性能.
- 为了解决保护医疗细分模型的未经探索的领域.
主要方法:
- 隐形标记微妙地调节模型不确定性,而不会改变细分输出.
- 不依赖模型的解释方法 (例如,LIME) 提取水印启示的特征属性.
- 一个QR码被设计成一个水印,用于强大的和可识别的所有权索赔.
主要成果:
- 隐形标记有效地验证了跨不同数据集和模型的医疗细分模型的所有权.
- 该方法是隐蔽和无害的,保持原始模型的性能 (例如,SAM模型的Dice/AUC下降<1%).
- 在保持细分精度的同时,实现了高的攻击成功率 (>95%),优于后门方法.
结论:
- 隐形Mark为医疗细分模型所有权验证提供了一种有效,隐形和无害的解决方案.
- 在QR码水印方便强大的和可识别的所有权索赔.
- 该方法显示了在保护有价值的医疗AI知识产权方面实际部署的强大潜力.
更多相关视频
相关概念视频
04:08The Uncertainty Principle
31.8K
Werner Heisenberg considered the limits of how accurately one can measure properties of an electron or other microscopic particles. He determined that there is a fundamental limit to how accurately one can measure both a particle’s position and its momentum simultaneously. The more accurate the measurement of the momentum of a particle is known, the less accurate the position at that time is known and vice versa. This is what is now called the Heisenberg uncertainty principle. He...
31.8K
03:00Self-Evaluation: Self-Enhancement and Self-Verification
5.8K
Social psychologists have documented that feeling good about ourselves and maintaining positive self-esteem is a powerful motivator of human behavior (Tavris & Aronson, 2008). In the United States, members of the predominant culture typically think very highly of themselves and view themselves as good people who are above average on many desirable traits (Ehrlinger, Gilovich, & Ross, 2005). Often, our behavior, attitudes, and beliefs are affected when we experience a threat to our...
5.8K
02:46Uncertainty in Measurement: Reading Instruments
51.6K
Counting is the type of measurement that is free from uncertainty, provided the number of objects being counted does not change during the process. Such measurements result in exact numbers. By counting the eggs in a carton, for instance, one can determine exactly how many eggs are there in the carton. Similarly, the numbers of defined quantities are also exact. For example, 1 foot is exactly 12 inches, 1 inch is exactly 2.54 centimeters, and 1 gram is exactly 0.001 kilograms. Quantities...
51.6K
01:17Strategies of Self-Presentation II: Self-Verification
189
Self-verification is a fundamental psychological drive wherein individuals seek affirmation of their self-concept from others, striving for consistency between their internal self-view and external perceptions. This drive operates even when the self-concept is negative, influencing interpersonal behavior and feedback preferences in complex and often counterintuitive ways. Unlike the self-enhancement motive, which seeks positive evaluations, self-verification prioritizes coherence and...
189
00:59Uncertainty: Overview
1.7K
In analytical chemistry, we often perform repetitive measurements to detect and minimize inaccuracies caused by both determinate and indeterminate errors. Despite the cares we take, the presence of random errors means that repeated measurements almost never have exactly the same magnitude. The collective difference between these measurements - observed values - and the estimated or expected value is called uncertainty. Uncertainty is conventionally written after the estimated or expected value.
1.7K
03:34Uncertainty in Measurement: Significant Figures
81.6K
All the digits in a measurement, including the uncertain last digit, are called significant figures or significant digits. Note that zero may be a measured value; for example, if a scale that shows weight to the nearest pound reads “140,” then the 1 (hundreds), 4 (tens), and 0 (ones) are all significant (measured) values.
81.6K