K^4: Online Log Anomaly Detection via Unsupervised Typicality Learning(arxiv.org)
arxiv.org
K^4: Online Log Anomaly Detection via Unsupervised Typicality Learning
https://arxiv.org/abs/2507.20051
1 comments
> Abstract: [...] (K^4) transforms arbitrary log embeddings into compact four-dimensional descriptors (Precision, Recall, Density, Coverage) using efficient k-nearest neighbor (k-NN) statistics. These descriptors enable lightweight detectors to accurately score anomalies without retraining. Using a more realistic online evaluation protocol, sets a new state-of-the-art (AUROC: 0.995-0.999), outperforming baselines by large margins while being orders of magnitude faster, with training under 4 seconds and inference as low as 4 us.