arXiv subjects

Joseph Pollock

Publications and source records attributed to Joseph Pollock.

2 recordsLinked to original sources

arXiv2025

Shape and spin state model of contact binary (388188) 2006 DP14 using combined radar and optical observations

Contact binaries are found throughout the solar system. The recent discovery of Selam, the satellite of MBA (152830) Dinkinesh, by the NASA LUCY mission has made it clear that the term `contact binary' covers a variety of different types of bi-modal mass distributions and formation mechanisms. Only by modelling more contact binaries can this population be properly understood. We determined a spin state and shape model for the Apollo group contact binary asteroid (388188) 2006 DP14 using ground-based optical and radar observations collected between 2014 and 2023. Radar delay-Doppler images and continuous wave spectra were collected over two days in February 2014, while 16 lightcurves in the Cousins R and SDSS-r filters were collected in 2014, 2022 and 2023. We modelled the spin state using convex inversion before using the SHAPE modelling software to include the radar observations in modelling concavities and the distinctive neck structure connecting the two lobes. We find a spin state with a period of $(5.7860\pm0.0001)$ hours and pole solution of $\lambda = (180\pm121)^\circ$ and $\beta = (-80\pm7)^\circ$ with morphology indicating a 520 m long bi-lobed shape. The model's asymmetrical bi-modal mass distribution resembles other small NEA contact binaries such as (85990) 1999 JV6 or (8567) 1996 HW1, which also feature a smaller `head' attached to a larger `body'. The final model features a crater on the larger lobe, similar to several other modelled contact binaries. The model's resolution is 25 m, comparable to that of the radar images used.

astro-ph.EP↗

arXiv2024

Free Record-Level Privacy Risk Evaluation Through Artifact-Based Methods

Membership inference attacks (MIAs) are widely used to empirically assess privacy risks in machine learning models, both providing model-level vulnerability metrics and identifying the most vulnerable training samples. State-of-the-art methods, however, require training hundreds of shadow models with the same architecture as the target model. This makes the computational cost of assessing the privacy of models prohibitive for many practical applications, particularly when used iteratively as part of the model development process and for large models. We propose a novel approach for identifying the training samples most vulnerable to membership inference attacks by analyzing artifacts naturally available during the training process. Our method, Loss Trace Interquartile Range (LT-IQR), analyzes per-sample loss trajectories collected during model training to identify high-risk samples without requiring any additional model training. Through experiments on standard benchmarks, we demonstrate that LT-IQR achieves 92% precision@k=1% in identifying the samples most vulnerable to state-of-the-art MIAs. This result holds across datasets and model architectures with LT-IQR outperforming both traditional vulnerability metrics, such as loss, and lightweight MIAs using few shadow models. We also show LT-IQR to accurately identify points vulnerable to multiple MIA methods and perform ablation studies. We believe LT-IQR enables model developers to identify vulnerable training samples, for free, as part of the model development process. Our results emphasize the potential of artifact-based methods to efficiently evaluate privacy risks.

cs.LG↗