One of the key advantages of Federated Learning (FL) is its ability to collaboratively train a Machine Learning (ML) model while keeping clients' data on-site. However, this can create a false sense of security. Despite not sharing private data increases the overall privacy, prior studies have shown that gradients exchanged during the FL training remain vulnerable to Gradient Inversion Attacks (GIAs). These attacks allow reconstructing the clients' local data, breaking the privacy promise of FL. GIAs can be launched by either a passive or an active server. In the latter case, a malicious server manipulates the global model to facilitate data reconstruction. While effective, earlier attacks falling under this category have been demonstrated to be detectable by clients, limiting their real-world applicability. Recently, novel active GIAs have emerged, claiming to be far stealthier than previous approaches. This work provides the first comprehensive analysis of these claims, investigating four state-of-the-art GIAs. We propose novel lightweight client-side detection techniques, based on statistically improbable weight structures and anomalous loss and gradient dynamics. Extensive evaluation across several configurations demonstrates that our methods enable clients to effectively detect active GIAs without any modifications to the FL training protocol.

Active microwave sensors, such as synthetic aperture radars (SARs), offer all-weather and daylight independent operability which is of great advantage for monitoring polar regions, where extreme environmental conditions and long period of darkness strongly limit other kinds of sensors. In addition, microwaves allow penetrating into dry snow and ice, making SAR measurements sensitive to the subsurface structure of glaciers and ice sheets. However, the retrieval of glacier subsurface parameters from SAR observations remains difficult due to their sensitivity to a large number of factors, including snow and ice properties, presence of layers, etc. The objective of this study is to attempt advancing the understanding of SAR measurements of glaciers and ice sheets. A combined analysis of polarimetric and tomographic measurements is carried out to derive a 3-D characterization of the scattering mechanisms occurring in a glacier subsurface scenario. The investigation exploits a fully-polarimetric tomographic airborne dataset, acquired over Greenland by the DLR's F-SAR system in the frame of the ARCTIC15 campaign.

Digital elevation models generated with SAR interferometry (InSAR) are an important information source for glacier and ice sheet mass balance. However, the measured elevations suffer from a penetration bias due to the interferometric phase center being up to several tens of meters below the surface. The penetration of the microwave signals depends on SAR parameters (e.g. frequency) and snow and ice conditions. There is potential to estimate this penetration bias directly from the data by means of polarimetric InSAR models. Existing models fail to describe the data across different test sites and ice conditions and phase centers were found to be deeper than predicted by these models. SAR tomography is employed to assess the vertical distribution of backscattering in the data from an airborne campaign. The data are compared to refined models in order to find better representations of the vertical backscattering distribution, while the model complexity is purposely kept simple to make a phase center estimation possible. Additionally, recent work showed the importance of strong subsurface layers which influence phase center depth. Combining the refined subsurface structure models and dominant subsurface layers allows simulating a variety of ice sheet subsurface scenarios and can be used to assess the potential to estimate the InSAR phase center depth directly from the data.

The need of large scale observations with high temporal frequency has promoted airborne and satellite remote sensing techniques for glaciological applications. In particular, active microwave sensors, such as synthetic aperture radars (SARs), offer all-weather and daylight independent operability which is of great advantage at high latitudes, where extreme environmental conditions and long period of darkness strongly limit other kinds of sensors. Moreover, longer wavelengths allow to penetrate significantly into dry snow and ice, interacting with surface as well as subsurface features. On the one hand, this makes SAR measurements suitable to investigate the subsurface structure of glaciers and ice sheets. On the other hand, the complex interaction of microwaves with the subsurface layers makes the interpretation of SAR measurements challenging. This study investigates the potential of SAR techniques to retrieve information about glacier subsurface. SAR polarimetry and tomography are used to gain a 3-D characterization of the scattering scenario of the K-transect, a site located in the ablation zone of Greenland. For this, a fully-polarimetric tomographic airborne dataset, acquired by the DLR's F-SAR system in the frame of the ARCTIC15 campaign, is exploited.

The vertical structure of firn below the surface of the Greenland ice sheet is driven by meteorological conditions like the intensity and duration of melting periods. The formation of ice inclusions through meltrefreeze processes leads to strong scatterers whose vertical distribution influences interferometric coherences. Interferometric coherence profiles from quad-pol SAR data show clear sensitivity to the different subsurfaces of two test sites with different melting periods. The differences in coherence are explained in this study through modelling of the different vertical subsurface structures supported by ground penetrating radar measurements. The shown sensitivity demonstrates the potential of polarimetric SAR interferometry to retrieve information about meltrefreeze processes and layering below the surface of ice sheets.

There is a long-stranding need of reliable space-borne observations on snow mass. Current satellite sensors and data products are largely unable to meet requirements presented in particular by numerical prediction and watershed management. Consequently, several concept studies have been initiated to address these specific needs, outlining possibilities for future space sensors focusing on retrieval of snow mass and other characteristics of the terrestrial cryosphere. The results of these of-going mission concept studies are presented and discussed. Several possible sensor options are presented, which would address diverse needs on either hemispheric or regional scales.

Spaceborne synthetic aperture radars (SARs) represent a powerful tool to perform cryospheric observations due to their high spatial resolution and capability to acquire data during the winter time. Especially at lower frequencies, SAR signals penetrate beneath the glacier surface through the shallow snow cover, interacting with surface as well as subsurface features. This makes the scattering scenario very complex and the interpretation of SAR backscattering from glaciers and ice sheets not straightforward. In the case of polarimetric SARs, the understanding of polarization phase differences represents one of the main open issues. In this letter, a physical model is employed to relate copolarization HH-VV phase difference (CPD) to structural and dielectric properties of snow and firn which characterize the uppermost layers of glaciers. Modelled CPD values are compared to values observed in airborne L-band (1.3 GHz) SAR data, acquired over the Austfonna ice cap, in Svalbard, in spring 2007. The inversion of the employed model shows promising results toward the retrieval of firn properties. In particular, the obtained thickness estimates are found to be in good agreement with the GPR profiles measured in coordination with the SAR acquisitions.

Over the last 30 years, the use of airborne and satellite remote sensing techniques has revolutionized glaciology through unequaled improvements in the scale and in the temporal and spatial resolutions of cryospheric observations. In this sense, the use of synthetic aperture radar (SAR) configurations likely was the greatest advance of the past two decades in the monitoring of the Earth's cryosphere. This paper describes a novel polarimetric scattering model to interpret polarimetric SAR (PolSAR) measurements of glaciers and ice sheets. Different scattering components are considered according to the specific glacier facies, in order to infer physical parameters of the ice volume. Total backscattering is modeled as the incoherent sum of possible surface and volume contributions. Surface scattering is described using the X-Bragg model in order to account for roughness, whereas clouds of spheroidal scatterers (either prolates or oblates) are considered to model volume scattering. The model includes differential propagation effects related to firn anisotropy and is therefore able to interpret polarimetric phase differences often present in PolSAR data acquired over glaciers. The developed model is used to interpret different scenarios imaged at L-band in fully polarimetric mode by the E-SAR sensor of the German Aerospace Center (DLR) over test sites located on the Austfonna ice cap, Svalbard, Norway, as part of the ICESAR2007 campaign. The obtained results are consistent with the reference information provided by the available ground measurements about the subsurface structure of the study area and encourage the use of dedicated scattering models for the different glacier zones.

In this study application results are presented derived from multi-parametric SAR observations covering five different thematic domains: forest, agriculture, ocean, urban and cryosphere. In total 21 application products have been selected and described. Their application on different data sets, space- and airborne sensors, was demonstrated and can independently be reproduced by any scientist. The results and algorithms are available soon through Springer.

The paper introduces a new polarimetric scattering model able to interpret and invert coherent polarimetric SAR (PolSAR) measurements over glaciers and ice sheets. Individual scattering components related to ice lenses and pipes are considered to model the subsurface scattering structure of ice sheets. The model is able to interpret the scattering amplitudes, their ratios at the different polarizations as well as the observed polarimetric phase differences. The co-polarization (HH-VV) phase difference is related to the structural anisotropy of the firn layer and will be used to estimate its thickness. The model is validated against L-band PolSAR data acquired by the E-SAR sensor of the German Aerospace Center (DLR) over the Austfonna ice cap in Svalbard during the ICESAR2007 campaign and available GPR profiles.

Spaceborne synthetic aperture radars (SARs) represent a powerful tool to perform cryospheric observations due to their high spatial resolution and capability to acquire data during the winter time. Especially at lower frequencies, SAR signals penetrate beneath the glaciers surface through the shallow snow cover, interacting with surface as well as sub-surface features. This makes the scattering scenario very complex and the interpretation of SAR backscattering from glaciers and ice sheets not straightforward. In the case of polarimetric SARs (PolSAR), the understanding of polarization phase differences represent one of the main open issues. In this paper, a physical model is employed to relate co-polarization HH-VV phase difference (co-pol phase) to structural and dielectric properties of snow and firn which characterize the first meters of glaciers subsurface. Modelled co-pol phase values are compared to values measured in experimental airborne L-band (1.3 GHz) SAR data, acquired over the Austfonna ice-cap, in Svalbard, in spring 2005 and 2007. A set of ground measurements is also exploited to properly set up the model and to validate the results. Finally, an interpretation of co-pol phase temporal variations over the period 2005-2007 is attempted based on the proposed model.

In recent years an increased interest arose in using synthetic aperture radar (SAR) to study and monitor land ice for glaciological applications and climate change research. At long wavelengths, SAR systems can penetrate into glacier ice for several tens to hundreds of meters. This makes them sensitive to near-surface as well as deeper ice volume features. This paper investigates the performance of a volume scattering component, in the course of an electromagnetic (e.m.) model development, to relate Polarimetric SAR (PolSAR) signatures to glacier facie. Indeed, PolSAR ice signatures are still poorly understood, including the importance of scattering from the glacier ice volume and its dependency on frequency and incidence angle. A comparison is performed with airborne Pol-SAR data at L- and P-band collected by DLR's E-SAR system over the Austfonna ice cap in Svalbard, Norway, within the ICESAR campaign.

The interest in using synthetic aperture radar (SAR) to study snow properties has increased in the last years. Snow properties and snow cover change monitoring on global scale are needed for a better understanding of the role of the cryosphere in the climate system [1]. Snow plays also an important role in biogeochemical cycles since it is a basic water resource for many densely populated areas of the planet. In this sense, Snow Water Equivalent (SWE) is the primary parameter to be retrieved. The Cold Regions Hydrology High-resolution Observatory (CoReH2O) satellite mission, selected by the ESA for feasibility studies within the Earth Explorer Programme, is expected to satisfy this need [2]. In this work, the capabilities of the proposed CoReH2O models for SWE (and snowflakes size) retrieval are investigated using TerraSAR-X dual-pol data. Ground measurements from the NoSREx campaign are used for validation.