David Kurtenbach, Kansas State University; Megan Manly, Kansas State University; Zach Metzinger, Kansas State University
Keywords: Autoencoder, Russia, Pattern of Life, Anomaly Detection, Deep Learning, SDA
Abstract:
We apply deep learning techniques for anomaly detection to analyze activity of Russian-owned resident space objects (RSO) prior to the Ukraine invasion and assess the results for any findings that can be used as indications or warnings of aggressive military behavior for future conflicts. This research looks to assess the existence of statistically significant changes in Russian RSO pattern of life/pattern of behavior (PoL/PoB) using Keplerian elements that are publicly available. This research looks at statistical and deep learning approaches to assess anomalous activity. The deep learning method uses and compares a variational autoencoder (VAE), traditional autoencoder (AE), an anchor loss based autoencoder (Anchor AE), isolation forest (IF), and Kolmogorov Aronold Network (KAN) approach to establish a baseline of on-orbit activity based on a five-year data sample. The primary investigation period focuses on the six months leading up to the invasion date of February 24, 2022. Additional analysis looks at RSO activity during an active combat period by sampling two-line element (TLE) data after the invasion date. The deep learning models identify anomalies based on reconstruction errors that surpass a threshold of two standard deviations. To capture the nuance and unique characteristics of each RSO an individual model was trained for each observed space object. The total number of autoencoder models trained and used for inference is 2,544 which is based on a select number of space objects that met the defined criteria to be included in the research. The research strived to prioritize explainability and interpretability of the model results thus each observation was assessed for anomalous behavior of the individual six orbital elements versus analyzing the input data as a single monolithic observation. The results demonstrate not just statistically significant anomalies of Russian RSO activity but detail anomalous findings to the individual Keplerian element. To drive explainability, objects were analyzed based on their categorized mission. This helps to create understandable results that can be extended to establish a generalized profile of space activity leading up to aggressive military actions. Codebase is available at https://github.com/davidkurtenb/Russat_AnomDetect
Date of Conference: September 16-19, 2025
Track: Machine Learning for SDA Applications