Yuming Xie

Yuhe Ji, Yilun Liu, Feiyu Yao et al.

Log analysis represents a critical sub-domain within AI applications that facilitates automatic approaches to fault and error management of large-scaled software systems, saving labors of traditional manual methods. While existing solutions using large language models (LLMs) show promise, they are limited by a significant domain gap between natural and log languages (the latter contains rich domain-specific tokens such as status codes, IP addresses, resource pathes), which restricts their effectiveness in real-world applications. However, directly adapting general-purpose LLMs to log analysis using raw logs may degrade their performance due to inconsistent token distribution. In this paper, we present a domain adaptation approach that addresses these limitations by integrating interpretable domain knowledge into open-source LLMs through continual pre-training (CPT), which bridges this domain gap by adapting LLMs on interpretable natural texts with log knowledge (instead of raw logs) to reduce distribution discrepancy. To achieve this, we developed NLPLog, a comprehensive dataset containing over 250,000 question-answer pairs on log-related knowledge. Our resulting model, SuperLog, achieves the best performance across four log analysis tasks, with an average accuracy improvement of 12.01% over the second-best model. Ablation study also suggests advantages of domain adaption using interpretable log knowledge over using raw logs.

Siyuan Guo, Huiwu Liu, Xiaolong Chen et al.

In this work, we explore the potential of large language models (LLMs) for generating functional test scripts, which necessitates understanding the dynamically evolving code structure of the target software. To achieve this, we propose a case-based reasoning (CBR) system utilizing a 4R cycle (i.e., retrieve, reuse, revise, and retain), which maintains and leverages a case bank of test intent descriptions and corresponding test scripts to facilitate LLMs for test script generation. To improve user experience further, we introduce Re4, an optimization method for the CBR system, comprising reranking-based retrieval finetuning and reinforced reuse finetuning. Specifically, we first identify positive examples with high semantic and script similarity, providing reliable pseudo-labels for finetuning the retriever model without costly labeling. Then, we apply supervised finetuning, followed by a reinforcement learning finetuning stage, to align LLMs with our production scenarios, ensuring the faithful reuse of retrieved cases. Extensive experimental results on two product development units from Huawei Datacom demonstrate the superiority of the proposed CBR+Re4. Notably, we also show that the proposed Re4 method can help alleviate the repetitive generation issues with LLMs.

Yangde Wang, Weidong Qiu, Yuming Xie et al.

Pattern lock is a general technique used to realize identity authentication and access authorization on mobile terminal devices such as Android platform devices, but it is vulnerable to the attack proposed by recent researches that exploit information leaked by users while drawing patterns. However, the existing attacks on pattern lock are environmentally sensitive, and rely heavily on manual work, which constrains the practicability of these attack approaches. To attain a more practical attack, this paper designs the PatternMonitor, a whole pipeline with a much higher level of automation system againsts pattern lock, which extracts the guessed candidate patterns from a video containing pattern drawing: instead of manually cutting the target video and setting thresholds, it first employs recognition models to locate the target phone and keypoints of pattern drawing hand, which enables the gesture can be recognized even when the fingertips are shaded. Then, we extract the frames from the video where the drawing starts and ends. These pre-processed frames are inputs of target tracking model to generate trajectories, and further transformed into possible candidate patterns by performing our designed algorithm. To the best of our knowledge, our work is the first attack system to generate candidate patterns by only relying on hand movement instead of accurate fingertips capture. The experimental results demonstrates that our work is as accurate as previous work, which gives more than 90\% success rate within 20 attempts.