Latent Sculpting for Zero-Shot Generalization: A Manifold Learning Approach to Out-of-Distribution Anomaly Detection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chhetri, Rajeeb Thapa, Thapa, Saurab, Kumar, Avinash, Chen, Zhixiong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911497779675136
author Chhetri, Rajeeb Thapa
Thapa, Saurab
Kumar, Avinash
Chen, Zhixiong
author_facet Chhetri, Rajeeb Thapa
Thapa, Saurab
Kumar, Avinash
Chen, Zhixiong
contents A critical vulnerability of supervised deep learning in high-dimensional tabular domains is "generalization collapse": models form precise decision boundaries around known training distributions but fail catastrophically when encountering Out-of-Distribution (OOD) data. To overcome this, we propose Latent Sculpting, a hierarchical, two-stage representation learning architecture designed to enforce explicit structural boundaries prior to density estimation. In the first stage, a Transformer-based tabular encoder is trained using our novel Binary Latent Sculpting loss. This objective explicitly condenses benign network traffic into a dense, low-entropy hypersphere while enforcing a strict geometric minimum-distance margin for anomalous patterns. In the second stage, a Masked Autoregressive Flow (MAF) maps this structurally optimized manifold to calculate exact, probabilistic anomaly thresholds. We evaluate this methodology on the CIC-IDS-2017 benchmark under a rigorous zero-shot protocol, deliberately withholding complex attack classes during training to test true OOD generalization. Averaged across three random initialization seeds to ensure statistical robustness, our framework maintains near-perfect classification on known signatures (F1 = 0.980 +/- 0.000) while achieving an overall zero-shot OOD F1-Score of 0.867 +/- 0.021 and an AUROC of 0.913 +/- 0.010 at an 85th-percentile threshold. Most notably, the model achieves an average recall of 78.7% (peaking at 97.2%) on stealthy "Infiltration" attacks and over 94% on low-volume DoS variations - complex distributional shifts where standard supervised and unsupervised baselines historically suffer near-total detection failure. These empirical results demonstrate that explicitly decoupling topological manifold structuring from probabilistic density estimation establishes a highly stable and scalable defense against zero-day cyber threats.
format Preprint
id arxiv_https___arxiv_org_abs_2512_22179
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Latent Sculpting for Zero-Shot Generalization: A Manifold Learning Approach to Out-of-Distribution Anomaly Detection
Chhetri, Rajeeb Thapa
Thapa, Saurab
Kumar, Avinash
Chen, Zhixiong
Machine Learning
Cryptography and Security
68T07, 62H30
I.2.6; C.2.0
A critical vulnerability of supervised deep learning in high-dimensional tabular domains is "generalization collapse": models form precise decision boundaries around known training distributions but fail catastrophically when encountering Out-of-Distribution (OOD) data. To overcome this, we propose Latent Sculpting, a hierarchical, two-stage representation learning architecture designed to enforce explicit structural boundaries prior to density estimation. In the first stage, a Transformer-based tabular encoder is trained using our novel Binary Latent Sculpting loss. This objective explicitly condenses benign network traffic into a dense, low-entropy hypersphere while enforcing a strict geometric minimum-distance margin for anomalous patterns. In the second stage, a Masked Autoregressive Flow (MAF) maps this structurally optimized manifold to calculate exact, probabilistic anomaly thresholds. We evaluate this methodology on the CIC-IDS-2017 benchmark under a rigorous zero-shot protocol, deliberately withholding complex attack classes during training to test true OOD generalization. Averaged across three random initialization seeds to ensure statistical robustness, our framework maintains near-perfect classification on known signatures (F1 = 0.980 +/- 0.000) while achieving an overall zero-shot OOD F1-Score of 0.867 +/- 0.021 and an AUROC of 0.913 +/- 0.010 at an 85th-percentile threshold. Most notably, the model achieves an average recall of 78.7% (peaking at 97.2%) on stealthy "Infiltration" attacks and over 94% on low-volume DoS variations - complex distributional shifts where standard supervised and unsupervised baselines historically suffer near-total detection failure. These empirical results demonstrate that explicitly decoupling topological manifold structuring from probabilistic density estimation establishes a highly stable and scalable defense against zero-day cyber threats.
title Latent Sculpting for Zero-Shot Generalization: A Manifold Learning Approach to Out-of-Distribution Anomaly Detection
topic Machine Learning
Cryptography and Security
68T07, 62H30
I.2.6; C.2.0
url https://arxiv.org/abs/2512.22179