标题： 从少标签到零标签：一种基于元学习的跨系统日志异常检测方法 显示英文标题

标题： From Few-Label to Zero-Label: An Approach for Cross-System Log-Based Anomaly Detection with Meta-Learning

摘要： 日志异常检测在确保软件系统的稳定性和可靠性方面起着关键作用。 然而，现有方法依赖于大量标记的日志数据，在实际应用中带来了重大挑战。 为了解决这个问题，跨系统迁移已被确定为一个关键的研究方向。 最先进的跨系统方法仅使用目标系统中的少量标记数据即可实现有前景的性能。 然而，它们对标记的目标日志的依赖性使得当标记日志不足时容易受到冷启动问题的影响。 为了克服这一限制，我们探索了一个新颖但研究较少的场景：零标记跨系统日志异常检测，其中目标系统日志完全未被标记。 为此，我们提出了FreeLog，一种与系统无关的表示元学习方法，消除了对标记目标系统日志的需求，在零标记条件下实现了跨系统日志异常检测。 在三个公共日志数据集上的实验结果表明，FreeLog在性能上可与依赖目标系统少量标记数据的最先进方法相媲美。 显示英文摘要

摘要： Log anomaly detection plays a critical role in ensuring the stability and reliability of software systems. However, existing approaches rely on large amounts of labeled log data, which poses significant challenges in real-world applications. To address this issue, cross-system transfer has been identified as a key research direction. State-of-the-art cross-system approaches achieve promising performance with only a few labels from the target system. However, their reliance on labeled target logs makes them susceptible to the cold-start problem when labeled logs are insufficient. To overcome this limitation, we explore a novel yet underexplored setting: zero-label cross-system log anomaly detection, where the target system logs are entirely unlabeled. To this end, we propose FreeLog, a system-agnostic representation meta-learning method that eliminates the need for labeled target system logs, enabling cross-system log anomaly detection under zero-label conditions. Experimental results on three public log datasets demonstrate that FreeLog achieves performance comparable to state-of-the-art methods that rely on a small amount of labeled data from the target system.