Aug 2026· Frontiers in Computing and Intelligent Systems· 0 citations· 18 references
TL;DR
The next big jump will not be driven by larger parameters, but smaller, more visible and dynamic systems that can be trusted by security analysts, as well as a disconnect between lab and reality.
Abstract
Web services are now pervasive in almost all industries, but this has made them a major target for hackers. Traditional signature-based solutions are no longer effective against zero‑day and polymorphic threats that evolve rapidly. The much heralded game‑changer is artificial intelligence (AI) particularly deep learning and more recently large language models. But the literature is fragmented: CNNs learn local features, Transformers capture long-range dependencies, and LLMs offer semantic understanding - yet each suffers from real-world limitations on interpretability, efficiency and security. In this survey we chart this progress and suggest that it's not model accuracy but trustworthiness that matters most. We carve up benchmark datasets (CSIC2010, CICIDS2017) and see a disconnect between lab and reality. Our message: the next big jump will not be driven by larger parameters, but smaller, more visible and dynamic systems that can be trusted by security analysts.
The rapid evolution of phishing, malware distribution, and other web-based attacks has made malicious Uniform Resource Locator (URL) detection an important cybersecurity research problem. Traditional blacklist and rule-based mechanisms provide efficient protection against known malicious addresses but may fail when att...
Ashwini B. Sable, Vijay More· International Journal of Lat...· 0 citations
Generative artificial intelligence has altered the operational character of phishing. Large language models can produce fluent, context-sensitive and highly personalised messages that resemble routine organisational correspondence, weakening filters that depend heavily on spelling errors, suspicious keywords or previou...
Santanu Jana· International Journal For Mu...· 0 citations
Across the datasets tested, this hybrid, explainability-aware approach consistently outperformed conventional intrusion detection baselines on prediction accuracy, detection capability and adaptability to new attack types, positioning it as a scalable model for real-time attack prediction and security analysis in moder...
Lalith Nivas Yadlapalli, C. N., S. Chintalapudi· International Journal of Inn...· 0 citations
One lesson emerges from the experiments: putting the effort into how traffic is written down, instead of making the classifier heavier, offers an economical and workable path to intrusion detection across heterogeneous network environments.
Asmaa Benchama, Khalid Zebbara· EPJ Web of Conferences· 0 citations
This work contributes a comparative evaluation of lightweight detection models, consistent attention to security-critical metrics, and interpretable insights to support practical cybersecurity deployment.
Nwagbara Chisom Telvin, Gilbert Imuetin Osaze Aimufua, Raymond Ternenge Igbudu· FUDMA Journal of Sciences· 0 citations
Web application security faces significant threats from SQL injection (SQLi) and Cross-Site Scripting (XSS) attacks, which are characterized by high variability and concealment. Traditional detection methods relying on rule matching or shallow machine learning features struggle to identify novel and obfuscated attacks....
Rong Liu· International Conference on...· 0 citations
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
Microsoft Research Blog· microsoft.comOct 12, 2026
StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics. The post Simulating student learning for better AI tutoring appeared first on Microsoft Research.
