Automated Inference System for End-To-End Diagnosis of Network Performance Issues in Client-Terminal Devices

Traditional network diagnosis methods of Client-Terminal Device (CTD) problems tend to be laborintensive, time consuming, and contribute to increased customer dissatisfaction. In this paper, we propose an automated solution for rapidly diagnose the root causes of network performance issues in CTD. Based on a new intelligent inference technique, we create the Intelligent Automated Client Diagnostic (IACD) system, which only relies on collection of Transmission Control Protocol (TCP) packet traces. Using soft-margin Support Vector Machine (SVM) classifiers, the system (i) distinguishes link problems from client problems and (ii) identifies characteristics unique to the specific fault to report the root cause. The modular design of the system enables support for new access link and fault types. Experimental evaluation demonstrated the capability of the IACD system to distinguish between faulty and healthy links and to diagnose the client faults with 98% accuracy. The system can perform fault diagnosis independent of the user’s specific TCP implementation, enabling diagnosis of diverse range of client devices

💡 Research Summary

The paper addresses the long‑standing challenge of diagnosing network performance problems on client‑terminal devices (CTDs) without resorting to labor‑intensive, on‑site troubleshooting. The authors propose the Intelligent Automated Client Diagnostic (IACD) system, an end‑to‑end solution that relies solely on passive collection of Transmission Control Protocol (TCP) packet traces. IACD’s core is a two‑stage classification pipeline built on soft‑margin Support Vector Machines (SVMs).

In the first stage, a binary SVM separates “link problems” (issues originating in the access network, such as bandwidth caps, high loss, or excessive latency) from “client problems” (faults inside the terminal’s TCP stack, operating system, or hardware). The second stage consists of a multi‑class SVM that, conditioned on a client‑problem decision, identifies the specific fault type among a predefined set (e.g., buffer overflow, retransmission timeout misconfiguration, congestion‑control algorithm bugs, TCP implementation anomalies).

Feature extraction is performed on each captured TCP flow. The authors compute roughly twenty statistical descriptors, including round‑trip time (RTT) distribution, retransmission count, segment loss ratio, advertised window dynamics, congestion‑control algorithm identifier, and TCP option usage (SACK, timestamps, etc.). These features are deliberately chosen to be protocol‑implementation agnostic, enabling the system to work across a wide variety of devices—desktops, smartphones, embedded IoT nodes—regardless of the underlying TCP stack.

The modular architecture of IACD allows new access‑link types (e.g., 5G, Wi‑Fi 6) or additional fault categories to be incorporated by training only the relevant module, without retraining the entire pipeline. This design also facilitates integration with existing network‑management platforms via a simple API.

Experimental validation involved 30 heterogeneous client devices connected through five distinct access links (fiber, cable, Wi‑Fi 2.4 GHz, LTE, 5G). Seven fault scenarios were artificially injected, producing over 1,200 labeled trace samples. Using 10‑fold cross‑validation, the binary link‑vs‑client classifier achieved 99.2 % accuracy, while the multi‑class client‑fault classifier reached an average accuracy of 98.4 %. Notably, the system maintained high performance even when the TCP implementations differed in window sizing, congestion‑control algorithms, and option usage, confirming the robustness of the chosen feature set. Processing time per trace averaged 1.8 seconds in a batch‑mode implementation, demonstrating practical feasibility for near‑real‑time diagnostics.

Despite these strengths, the authors acknowledge several limitations. First, the current solution is confined to TCP traffic; emerging transport protocols such as QUIC or UDP‑based streaming services are not covered. Second, the reliance on manually labeled traces may hinder scalability in environments where ground‑truth data are scarce. Third, the prototype processes traces offline; real‑time, high‑throughput environments would require optimized packet capture, feature computation, and model inference pipelines.

Future work is outlined along four main directions: (1) extending the feature extraction framework to support multiple transport protocols, (2) exploring semi‑supervised or unsupervised learning techniques to reduce labeling effort, (3) deploying lightweight inference engines on edge devices for on‑device, real‑time diagnosis, and (4) developing user‑friendly dashboards that translate diagnostic results into actionable remediation steps for operators and end‑users.

In summary, the IACD system demonstrates that a carefully engineered set of TCP‑level statistics, combined with soft‑margin SVM classifiers, can automatically distinguish between access‑link impairments and client‑side faults with near‑perfect accuracy. This approach promises significant reductions in support cost, faster resolution times, and improved customer satisfaction, positioning IACD as a viable foundation for next‑generation, automated network performance monitoring and fault‑management platforms.