Global ETD Search
Search theses and dissertations gathered from participating repositories worldwide. Every result links back to the library that holds it. No account is needed.
Results
Showing 1 to 8 of 8 for “"Gradient Inversion"”.
-
Images in motion?: a first look into video leakage in federated learning
… only exchange model updates, such as gradients, and never their raw, sensitive data. This approach is fundamental for applications in domains where privacy and confidentiality are important. However, the security of this very mechanism is threatened by gradient inversion attacks, which …
-
Integrated interpretation of the Blake River Group using geophysical data from metal earth's Rouyn-Noranda transect, Quebec
… and unconstrained models. A 2-D conjugate gradient inversion of Magnetotelluric data was done after the phase tensor and skew angle were analyzed to determine dimensionality and directionality. The data was interpreted together in the form of integrated interpretation in order to understand …
-
LOW-POWER AND SECURE IMPLEMENTATION OF NEURAL NETWORKS BY APPROXIMATION
… More specifically, we consider the Model Inversion Attacks (MIAs) that extrapolate the training data from model parameters. Our proposed solution -- MIDAS: Model Inversion Defenses with an Approximate memory System -- intentionally introduces memory faults by overscaling voltage to thwart …
-
Understanding and measuring flow in aortic stenosis with MRI
… the duration of the velocity sensitivity gradients and shortening the overall echo time (TE). However, improvements in an optimised PC sequence (TE 1:5ms) were not satisfactory. Flow estimates remained variable and were underestimated beyond the aortic valve. To reduce the TE further, a …
-
Modeling and inversion of stress-induced multicomponent seismic time shifts
… for large values of [Delta]P. A hybrid global/gradient inversion algorithm requiring neither Jacobian or Hessian computations is developed to invert noise-contaminated time shifts for reservoir pressure and length. Numerical testing shows that P-wave shifts from the top of the reservoir produce …
-
Breaking Privacy in Model-Heterogeneous Federated Learning
… instead, clients only share locally computed gradients or model parameters with a central server. As individual gradients may leak information about a given client's dataset, secure aggregation was proposed. With secure aggregation, the server only receives the aggregate gradient update from …
-
Towards Personalized, Privacy-preserving, and Explainable AI for Healthcare
… we suggest a robust defense method against gradient inversion attacks on medical image data via latent space data perturbation and minimax optimization in Chapter 5.</p><p>We have summarized the significance and unresolved challenges in remote health monitoring. However, it’s important to …
-
On the solution of discrete-valued inverse problems through guided fuzzy c-means clustering
Geophysical inversion is a valuable tool for construction and interpretation of subsurface physical property models. Traditional inverse formulations recover smooth models, hindering interpretation of different units. Constrained inversions present methods to incorporate additional information and …