Search for: All records

Machine learning models are vulnerable to both security attacks (e.g., adversarial examples) and privacy attacks (e.g., private attribute inference). We take the first step to mitigate both the security and privacy attacks, and maintain task utility as well. Particularly, we propose an information-theoretic framework to achieve the goals through the lens of representation learning, i.e., learning representations that are robust to both adversarial examples and attribute inference adversaries. We also derive novel theoretical results under our framework, e.g., the inherent trade-off between adversarial robustness/utility and attribute privacy, and guaranteed attribute privacy leakage against attribute inference adversaries. 

Mechanical stretch can activate long-lived changes in fibroblasts, increasing their contractility and initiating phenotypic transformations. This activation, critical to wound healing and procedures such as skin grafting, increases with mechanical stimulus for cells cultured in two-dimensional but is highly variable in cells in three-dimensional (3D) tissue. Here, we show that static mechanical stretch of cells in 3D tissues can either increase or decrease fibroblast activation depending upon recursive cell–extracellular matrix (ECM) feedback and demonstrate control of this activation through integrated in vitro and mathematical models. ECM viscoelasticity, signaling dynamics, and cell mechanics combine to yield a predictable, but nonmonotonic, relationship between mechanical stretch and long-term cell activation. Results demonstrate that feedback between cells and ECM determine how cells retain memory of mechanical stretch and have direct implications for improving outcomes in skin grafting procedures. 

In Proceedings of the 33rd USENIX Security Symposium (USENIX Security), Philadelphia, PA, USA, August 14-16, 2024.