# METANOIA: A Lifelong Intrusion Detection and Investigation System for Mitigating Concept Drift

Source: [METANOIA: A Lifelong Intrusion Detection and Investigation System for Mitigating Concept Drift](<https://www.sota2.com/research/paper/68fa766d1e209131cea6ce55>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-03-19T05:00:05.030Z

Authors: Jie Ying, Mengce Zheng, Jungan Chen, Ruoxi Chen, Zhongjie Zhua, Tiantian Zhu

Published: 2024-12-31T13:44:27.000Z

## Abstract

As Advanced Persistent Threat (APT) complexity increases, provenance data is increasingly used for detection. Anomaly-based systems are gaining attention due to their attack-knowledge-agnostic nature and ability to counter zero-day vulnerabilities. However, traditional detection paradigms, which train on offline, limited-size data, often overlook concept drift - unpredictable changes in streaming data distribution over time. This leads to high false positive rates. We propose incremental learning as a new paradigm to mitigate this issue. However, we identify FOUR CHALLENGES while integrating incremental learning as a new paradigm. First, the long-running incremental system must combat catastrophic forgetting (C1) and avoid learning malicious behaviors (C2). Then, the system needs to achieve precise alerts (C3) and reconstruct attack scenarios (C4). We present METANOIA, the first lifelong detection system that mitigates the high false positives due to concept drift. It connects pseudo edges to combat catastrophic forgetting, transfers suspicious states to avoid learning malicious behaviors, filters nodes at the path-level to achieve precise alerts, and constructs mini-graphs to reconstruct attack scenarios. Using state-of-the-art benchmarks, we demonstrate that METANOIA improves precision performance at the window-level, graph-level, and node-level by 30%, 54%, and 29%, respectively, compared to previous approaches.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Graph-level Attack Detection on E3-CADETS](<https://www.sota2.com/research/sota/graph-level-attack-detection-on-e3-cadets.md>): Graph-level attack detection; E3-CADETS; Precision: 80.

- [Graph-level attack detection on E3-THEIA](<https://www.sota2.com/research/sota/graph-level-attack-detection-on-e3-theia.md>): Graph-level attack detection; E3-THEIA; Precision: 100.

- [Anomaly Detection on E3-CADETS](<https://www.sota2.com/research/sota/anomaly-detection-on-e3-cadets.md>): Anomaly Detection; E3-CADETS; True Positives (TP): 11.

- [Anomaly Detection on E3-ClearScope](<https://www.sota2.com/research/sota/anomaly-detection-on-e3-clearscope.md>): Anomaly Detection; E3-ClearScope; True Positives: 2.

- [Anomaly Detection on THEIA E3](<https://www.sota2.com/research/sota/anomaly-detection-on-theia-e3.md>): Anomaly Detection; THEIA E3; True Positives (TP): 9.

- [Anomaly Detection on E5-THEIA](<https://www.sota2.com/research/sota/anomaly-detection-on-e5-theia.md>): Anomaly Detection; E5-THEIA; True Positives (TP): 2.

- [Node-level Anomaly Detection on DARPA TC CADETS E3](<https://www.sota2.com/research/sota/node-level-anomaly-detection-on-darpa-tc-cadets-e3.md>): Node-level Anomaly Detection; DARPA TC CADETS E3; Precision: 100.

- [Node-level Anomaly Detection on DARPA TC E3-THEIA](<https://www.sota2.com/research/sota/node-level-anomaly-detection-on-darpa-tc-e3-theia.md>): Node-level Anomaly Detection; DARPA TC E3-THEIA; Precision: 100.

- [Node-level Anomaly Detection on DARPA TC CADETS E5](<https://www.sota2.com/research/sota/node-level-anomaly-detection-on-darpa-tc-cadets-e5.md>): Node-level Anomaly Detection; DARPA TC CADETS E5; Precision: 100.

- [Node-level Anomaly Detection on DARPA TC E5-THEIA](<https://www.sota2.com/research/sota/node-level-anomaly-detection-on-darpa-tc-e5-theia.md>): Node-level Anomaly Detection; DARPA TC E5-THEIA; Precision: 100.

- [Graph-level attack detection on E3-ClearScope](<https://www.sota2.com/research/sota/graph-level-attack-detection-on-e3-clearscope.md>): Graph-level attack detection; E3-ClearScope; Precision: 100.

- [Anomaly Detection on E5-CADETS](<https://www.sota2.com/research/sota/anomaly-detection-on-e5-cadets.md>): Anomaly Detection; E5-CADETS; True Positives (TP): 6.

- [Anomaly Detection on E5-ClearScope](<https://www.sota2.com/research/sota/anomaly-detection-on-e5-clearscope.md>): Anomaly Detection; E5-ClearScope; TP: 3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2501.00438>)

[PDF](<https://arxiv.org/pdf/2501.00438.pdf>)
