Zijia Song

Zijia Song

Auto-generated profile, based on publicly available information

Zijia Song

National University of Defense Technology, Hefei, Anhui, China

Research Focus & Metrics

Academic Career Length

1Year (2026-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
16 September 2026
Verified
Book Chapter
Closed access
PDF file
YWYelin Wang
Yelin Wang
ZSZijia Song
Zijia Song
SYShuo Ye
Shuo Ye
CYChuanguang Yang
Chuanguang Yang
MWMiaoyu Wang
Miaoyu Wang
ZYZitong Yu
Zitong Yu

15 July 2026 in SSRN Electronic Journal
Verified
Preprint
Open access
DQDongzhen Qi
Dongzhen Qi
PCPan Chen
Pan Chen
ZSZijia Song
Zijia Song
XZXinyun Zhao
Xinyun Zhao
QYQiming Yu
Qiming Yu
GYGuozheng Yang
Guozheng Yang
BGP prefix hijacking compromises Internet routing reachability and may lead to large-scale traffic misdirection and service disruption, making it a critical security threat to the stable operation of the backbone Internet. Although mechanisms such as RPKI have reduced this risk, eliminating hijacking remains difficult because of practical deployment and availability trade-offs. Existing detection methods mainly focus on individual hijacking events and provide limited insight into which autonomous systems (ASes) are more likely to become hijackers. To address this gap, we construct positive and negative AS observation-window samples from multi-source public routing data and extract structured features from AS-PATH changes and MOAS states, together with image features from historical prefix-announcement behavior. We then propose PHD-RG, a multimodal fusion framework for discovering potential BGP prefix-hijacker ASes. PHD-RG uses structured features as the primary discriminative modality and introduces image-based information through gated residual correction, enabling end-to-endclassification. ExperimentalresultsshowthatPHD-RGwiththeSwin T image encoder achieves strong overall performance, reaching an AUC of 98.40% and an accuracy of 93.42%. In an operational screening scenario where hijacker instances account for only 2% of the candidate pool, PHD-RG needs to inspect only the top 5% of risk-ranked ASes to cover 91.0% of hijacker instances, reducing the verification workload by 95.0%.Theseresultsdemonstrateitspractical value for routing-security risk verification.

10 April 2026 in Computer Networks
Verified
Journal Article
Open access
ZSZijia Song
Zijia Song
ZMZhengyi Ma
Zhengyi Ma
QYQiming Yu
Qiming Yu
XXXiaohui Xie
Xiaohui Xie
YWYelin Wang
Yelin Wang
GYGuozheng Yang
Guozheng Yang



14 November 2026
Verified
Conference Paper
Open access
ZCZe Chen
Ze Chen
QYQiming Yu
Qiming Yu
ZSZijia Song
Zijia Song
GYGuozheng Yang
Guozheng Yang
WYWei Yan
Wei Yan
The widespread adoption of encrypted traffic poses severe challenges to current security situational awareness systems based on network traffic monitoring. In existing dataset-driven training and testing studies, limitations such as shortcut learning induced by spurious feature correlations and sample imbalance caused by the long-tail distribution of real-world traffic result in weak generalization of traffic identification performance to real-world network traffic. To address these limitations, we propose TDDM-Melatt. First, we design Melatt, a memory-decoupled traffic representation model, which employs Competitive Gating Long Short-Term Memory (CG-LSTM) to construct the encoder and decoder. We design a spurious-correlation-free pre-training and inference paradigm, employing strict topology anonymization and a frozen pre-trained encoder strategy to cut off the model's learning pathways for spurious features. Second, we propose a Traffic Denoising Diffusion Model (TDDM) tailored to the characteristics of traffic data. Addressing the structured nature of high-dimensional sparsity and strong feature correlations in traffic data, we solve the problem of having only class labels as conditional guidance for traffic diffusion models. Extensive experiments are conducted on 4 representative public benchmark datasets. Under strict flow-level splitting and anonymization, TDDM-Melatt outperforms 6 basic classification models and 6 SOTA representation learning models. The proposed method provides a new and effective technical pathway for encrypted traffic classification in real-world network environments.