# From Zero-Shot Machine Learning to Zero-Day Attack Detection

arXiv: https://arxiv.org/abs/2109.14868 · HTML: https://mlsanity.com/paper/2109.14868/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 4 of its 41 references that are in the map.
- First posted on arXiv 30 Sep 2021.

## Builds on (most cited first, 4 of 4)

- [Zero-Shot Learning -- The Good, the Bad and the Ugly](https://mlsanity.com/paper/1703.04394/) (2017)
- [On Wasserstein Two Sample Testing and Related Families of Nonparametric Tests](https://mlsanity.com/paper/1509.02237/) (2015)
- [Benchmarking the Benchmark -- Analysis of Synthetic NIDS Datasets](https://mlsanity.com/paper/2104.09029/) (2021)
- [Generalised Zero-Shot Learning with Domain Classification in a Joint Semantic and Visual Space](https://mlsanity.com/paper/1908.04930/) (2019)

## Built on by (most cited first, 10 of 10)

- [On Generalisability of Machine Learning-based Network Intrusion Detection Systems](https://mlsanity.com/paper/2205.04112/) (2022)
- [Zero Day Threat Detection Using Graph and Flow Based Security Telemetry](https://mlsanity.com/paper/2205.02298/) (2022)
- [Neuro-Symbolic AI for Cybersecurity: State of the Art, Challenges, and Opportunities](https://mlsanity.com/paper/2509.06921/) (2025)
- [Zero Day Threat Detection Using Metric Learning Autoencoders](https://mlsanity.com/paper/2211.00441/) (2022)
- [DOC-NAD: A Hybrid Deep One-class Classifier for Network Anomaly Detection](https://mlsanity.com/paper/2212.07558/) (2022)
- [Stacking an autoencoder for feature selection of zero-day threats](https://mlsanity.com/paper/2311.00304/) (2023)
- [A Hierarchical IDS for Zero-Day Attack Detection in Internet of Medical Things Networks](https://mlsanity.com/paper/2508.10346/) (2025)
- [Adaptive Deception Framework with Behavioral Analysis for Enhanced Cybersecurity Defense](https://mlsanity.com/paper/2510.02424/) (2025)
- [A Comprehensive Study of Supervised Machine Learning Models for Zero-Day Attack Detection: Analyzing Performance on Imbalanced Data](https://mlsanity.com/paper/2512.07030/) (2025)
- [An empirical analysis of zero-day vulnerabilities disclosed by the zero day initiative](https://mlsanity.com/paper/2512.15803/) (2025)

## Abstract (arXiv)

The standard ML methodology assumes that the test samples are derived from a set of pre-observed classes used in the training phase. Where the model extracts and learns useful patterns to detect new data samples belonging to the same data classes. However, in certain applications such as Network Intrusion Detection Systems, it is challenging to obtain data samples for all attack classes that the model will most likely observe in production. ML-based NIDSs face new attack traffic known as zero-day attacks, that are not used in the training of the learning models due to their non-existence at the time. In this paper, a zero-shot learning methodology has been proposed to evaluate the ML model performance in the detection of zero-day attack scenarios. In the attribute learning stage, the ML models map the network data features to distinguish semantic attributes from known attack (seen) classes. In the inference stage, the models are evaluated in the detection of zero-day attack (unseen) classes by constructing the relationships between known attacks and zero-day attacks. A new metric is defined as Zero-day Detection Rate, which measures the effectiveness of the learning model in the inference stage. The results demonstrate that while the majority of the attack classes do not represent significant risks to organisations adopting an ML-based NIDS in a zero-day attack scenario. However, for certain attack groups identified in this paper, such systems are not effective in applying the learnt attributes of attack behaviour to detect them as malicious. Further Analysis was conducted using the Wasserstein Distance technique to measure how different such attacks are from other attack types used in the training of the ML model. The results demonstrate that sophisticated attacks with a low zero-day detection rate have a significantly distinct feature distribution compared to the other attack classes.