Loading article...
Loading article...
Stream/Area:
-
Subject:
-
Paper ID:
IJSREM69144
Volume:
10 |
11 |
Nov 2026
Pages:
1-7
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.
Computer Science
MIT ADT UNIVERSITY
Computer Science
MIT ADT UNIVERSITY
Computer Science
MIT ADT UNIVERSITY
Computer Science
MIT ADT UNIVERSITY
Computer Science
MIT ADT UNIVERSITY
Fake websites that imitate trusted brands to harvest credentials, most commonly phishing sites, are difficult to stop with conventional blocklists because campaigns rotate through disposable infrastructure within hours. This paper presents a dual-model detection architecture that achieves a 96.7% F1-score with a 1.8% false positive rate while maintaining sub-80 ms response latency at the 95th percentile and throughput exceeding 2,400 classifications per second. The system combines gradient boosting for rapid primary inference with a character-level neural network for refined secondary analysis of ambiguous cases. Feature extraction spans 82 discriminative signals across URL lexical patterns, DNS resolution metadata, TLS certificate properties, and DOM structural indicators. Confidence-based early termination routes approximately 82% of requests through the fast path, while asynchronous enrichment prevents network delays from blocking decisions. Temporal validation on chrono-logically partitioned datasets prevents the optimistic bias that arises from campaign overlap between training and test data. Ablation experiments quantify individual feature contributions, and systematic error analysis reveals misclassification patterns that inform future enhancements. The framework supports inline deployment across browser extensions, enterprise web gateways, and email security platforms.
KEYWORDS
This research demonstrates that effective fake and phish-ing website defense need not sacrifice operational viabil-ity. Through multi-modal signal integration, hybrid inference with confidence-based routing, and deployment-conscious op-timization, the proposed architecture achieves a 96.7% F1-score with a 1.8% false positive rate while sustaining more than 2,400 classifications per second at sub-80 ms tail latency. Temporal evaluation reveals realistic generalization char-acteristics that investigations using random data partitioning do not capture, and systematic ablation quantifies category contributions to guide deployment decisions under resource constraints. By bridging the persistent gap between research-grade accuracy and operational feasibility, this work provides a practical blueprint for next-generation real-time phishing de-fense infrastructure capable of protecting organizations against continuously evolving social-engineering threats.
[1] Anti-Phishing Working Group, “Phishing Activity Trends Report,” 2023–2024. [2] FBI Internet Crime Complaint Center, “Internet Crime Report,” 2024. [3] Verizon, “Data Breach Investigations Report,” 2024. [4] ENISA, “Threat Landscape Report,” 2023. [5] C. Seifert, I. Welch, and P. Komisarczuk, “Identification of malicious web pages with static heuristics,” in Proc. ATNAC, 2008. [6] Y. Zhang, J. Hong, and L. Cranor, “CANTINA: A content-based approach to detecting phishing web sites,” in Proc. WWW, 2007. [7] S. Abu-Nimeh, D. Nappa, X. Wang, and S. Nair, “A comparison of machine learning techniques for phishing detection,” in Proc. eCrime, 2007. [8] A. Oest, P. Zhang, B. Wardman, E. Nunes, J. Burgis, A. Zand, K. Thomas, A. Doupe´, and G.-J. Ahn, “Sunrise to sunset: Analyzing the end-to-end life cycle and effectiveness of phishing attacks at scale,” in Proc. USENIX Security, 2020. [9] J. Ma, L. Saul, S. Savage, and G. Voelker, “Beyond blacklists: Learning to detect malicious web sites from suspicious URLs,” in Proc. KDD, 2009. [10] C. Whittaker, B. Ryner, and M. Nazif, “Large-scale automatic classifi-cation of phishing pages,” in Proc. NDSS, 2010.
Publication Date
11/18/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
0
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...