Log Anomaly Detection by Leveraging LLM-Based Parsing and Embedding with Attention Mechanism
Bibliographic record
Abstract
During the software operation phase, automated log analysis is crucial for the early detection of anomalies to prevent critical incidents, like system failure. Learning-based anomaly detection techniques have shown the potential for real-time anomaly detection from trace logs through learning the execution patterns. However, extracting features from raw text format log files of diversified structures has been challenging and tackled in different ways. With the recent advancements in large language models (LLM), several LLM-based parsing methods have been proposed, where most of these methods struggled with uncertain output from LLM or manual rules set requirements for the parsing. To address these challenges, we have proposed a hybrid framework leveraging LLM in parsing and embedding. Our proposed approach uses the LLM to generate Regular expressions (REGEX) for the parser, along with parsing and event embedding (EM) using a pre-trained LLM model. Then, this framework leverages the reconstructive capacity of the autoencoder with attention mechanism (AM) for unsupervised learning of log patterns. The experimental case study shows the model’s effectiveness in anomaly detection using a public dataset with 96% accuracy. This framework will provide flexibility to pre-process different text-based log structures without human involvement in parsing.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.001 | 0.002 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.001 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".