MUMBAI, India, Oct. 26 -- Intellectual Property India has published a patent application (202641123374 A) filed by Ms. Ananthi. S. N; Ms. Gunanandhini S; Ms. Aswathy M P; and S. A. Engineering College on October 15, 2026, for Ai-Driven Prompt-Based Framework For Intelligent Anomaly Detection In System Logs And Network Traffic.
Inventors include Ms. Ananthi. S. N; Ms. Gunanandhini S; and Ms. Aswathy M P.
The application for the patent was published on October 23, 2026, under issue no. 43/2026.
Abstract: ABSTRACT Tracking, diagnosing, and guaranteeing the dependability of huge computer systems all depend on log data, therefore making it a priceless asset. Nevertheless, the lack of organization, tremendous power in the dimensionality and enormous size introduce significant challenges to the anomaly detection. Traditional machine learning models are often over reliant on manual preprocessing, manually engineered features and domain expertise, thus being less scaling and more flexible. The methods of deep learning also require huge amounts of training data that are labeled, costly and time consuming to obtain in practice, thereby restricting its strength. These thresholds hamper the detection of anomalies, which are crucial towards avoiding system crashes, ensuring service as well as achieving cybersecurity. One way to break these constraints and to analyze the log messages as human language, an anomaly detection system based on log prompts, which will rely on a pretrained language model and log-specifically designed prompts is proposed. This framework based on prompts extracts hidden contextual relationships and structural variations in data streams of logs, therefore it permits more effective anomaly detection with follow of this minimum preprocessing and limited labeled data. Scalable to distributed systems, the cloud infrastructures and enterprise level monitoring tools, the technique offers some flexibility to heterogeneous log structures and domains. Results on benchmark log datasets reveal that Performance wise, the framework surpasses both conventional machine learning and current deep learning standards. Specifically, the framework's strength was shown by the improved precision, recall, and F1-score as well as lower false positive rates that were noted. Combining the pre-training benefits of language models with domain-adaptive prompting answers for scalable, clear, and quick log anomaly detection in modern logs.
Disclaimer: Curated by HT Syndication.
