Contrastive Learning Algorithm for Low-Resource Cryptographic Attack Event Detection
The widespread use of the internet has led to frequent cryptographic attack event incidents, which pose various risks, including the leakage of personal information, privacy data, identity theft, and potential legal liabilities. However, conventional event detection models depend on expert-annotated...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
IEEE
2025-01-01
|
| Series: | IEEE Access |
| Subjects: | |
| Online Access: | https://ieeexplore.ieee.org/document/11045891/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The widespread use of the internet has led to frequent cryptographic attack event incidents, which pose various risks, including the leakage of personal information, privacy data, identity theft, and potential legal liabilities. However, conventional event detection models depend on expert-annotated trigger words, which are costly to obtain and constrain the applicability of detection methods in cryptographic security. Detecting cryptographic attack events faces two challenges: 1) Semantic complexity: terminology and abbreviations are diverse, and the same event is described using different vocabulary, sentence structures, and contexts, making it difficult for models to accurately identify specific events; 2) Lack of annotated features: This field relies on manually annotated data for feature learning, but there is a lack of publicly available datasets, and related research is scarce. Thus, we propose a method CLAD: Contrastive Learning Algorithm for Detecting Low-resource Cryptographic Attack Event. By comparing similarities and differences between samples, richer feature representations can be extracted. Additionally, assigning appropriate weights to keywords of different event types and employing clustering methods for effective classification reduce reliance on manually labeled data, thereby facilitating the training of an efficient cryptographic attack event detection model. To evaluate the effectiveness of the proposed method, we analyzed 916 cryptographic attack events collected from news blogs. The evaluation results show that the proposed method achieved an F1 score of 82.05% in detecting cryptographic attack events in low-resource scenarios, significantly outperforming state-of-the-art solutions. It reduces reliance on manual annotation and effectively handles semantic complexity using contrastive learning and clustering techniques. |
|---|---|
| ISSN: | 2169-3536 |