New traffic classes in seconds, without retraining
Classification by retrieval lets an encrypted-traffic model add new application classes without retraining, and flags out-of-distribution traffic in the same retrieval step.
Chen Hajaj
Machine learning that keeps working when data, threats, and domains change.
76 refereed journal and conference papers since 2012
Each square is a paper (33 journal, 43 conference), coloured by its main research line. Hollow squares are the 9 short and workshop papers and abstracts, and 12 preprints, theses, patents and books. Hover or tap one to read it.
In the Department of Industrial Engineering and Management I also head the Data Science track of the undergraduate and graduate programs. I chair the Faculty of Engineering Review Board and serve as an NVIDIA University Ambassador.
- Citations
- 1,025
- h-index
- 18
- Co-authors
- 128
- Students in the lab
- 20
Now
-
New paper in Computers & Security: HTTP2DoS: Detecting low-and-slow HTTP/2 DoS attacks with complex transfer-entropy graph-temporal learning, with Udi Aharon, Ran Dubin and Amit Dvir. Open access.
-
Paper accepted at IEEE CAMAD 2026: FALCON: Few-shot ANN for Low-latency Classification in O-RAN Networks — with Liav Cohen and Amit Dvir. Will be presented in the AI-Driven 6G Cloud, O-RAN and Network Orchestration session.
-
Best Paper Award (2nd Place) at the 20th ILAIS Conference 2026, Peres Academic Center: Product Pricing by Multidimensional Similarity: A Multimodal Attention Framework Using Image, Text, and Temporal Representations — with Eylon Yehiel and Anat Goldstein.
-
Congratulations to Yarin Amar for being named on the Rector’s List for academic excellence, and to Eylon Yehiel for being awarded the Dean’s List Excellence award. We are proud of your outstanding achievements!
-
Paper accepted at DCC 2026 (Design, Computing and Cognition), Paris: Learning through Concept Generation: The Case of Academic and Practitioner Teaching in the Design Engineering Studio.
Research
Six lines of work, each with its own page and the papers behind it. Some papers belong to two lines.
-
Adversarial Artificial Intelligence
Developing robust AI systems that withstand adversarial attacks — across network security, mobile malware, healthcare AI, and social network manipulation.
29 papers, 2018 to 2026
-
Encrypted Traffic Classification
AI-powered classification of encrypted network traffic — enabling security monitoring and QoS management without compromising encryption or user privacy.
20 papers, 2020 to 2027
-
Medical Data Science
Machine learning for critical care — predicting ICU feeding complications, phenotyping rare respiratory diseases, and enabling privacy-preserving medical data collaboration.
17 papers, 2015 to 2026
-
Incentive Design
Mechanism design and game theory for e-commerce platforms, team formation, and kidney exchange — aligning individual incentives with collective outcomes.
21 papers, 2013 to 2022
-
Quantifying Constructivist Learning in Studio-Based Education
Data science methods that quantify learning in design studio education — measuring cognitive breakthroughs, engagement, and teaching effectiveness during live critiques.
7 papers, 2021 to 2026
-
E-commerce and Multimodal AI
Multimodal AI that fuses computer vision and NLP to measure item similarity the way humans do, and studies of how people shop online across devices, for smarter recommendations in tourism, e-commerce, and real estate.
7 papers, 2020 to 2026
In plain words
Short explainers of what our papers mean in practice, each with one figure.
- Encrypted traffic When the test changes the winner Papers on traffic classification usually credit their gains to a newer model. We trained six deep models 2,016 times and found that how the data is...
- E-commerce and multimodal AI Recommending a product nobody has clicked yet Recommendation engines learn from clicks and purchases, so a brand-new product has nothing to learn from. We describe a new item by its photo, its text...
- Design education How a design critique sounds In the design studio, students learn in critiques: short conversations with a tutor about their work. With Hadas Sopher and John Gero, we use language models...
From our lab
Recent papers led by Chen or a lab student, or written by a small team.
-
When Evaluation Protocol Changes the Winner: Traffic Classification Beyond Architecture
-
Cleaner Adversarial CAPTCHAs: Intelligent Targets and Precise Noise for Usable Security
-
Product Pricing by Multidimensional Similarity: A Multimodal Attention Framework Using Image, Text, and Temporal Representations
-
Ultra fast encrypted network traffic classification based on classification by retrieval powered by out-of-distribution
-
Machine Learning Tools for Predicting Pediatric Urinary Tract Infections Caused by ESBL-Producing Bacteria
-
Leveraging OSINT for Advanced Proactive Cybersecurity: Strategies and Solutions
The lab
9 PhD and 11 MSc students work in the lab today. 3 PhD and 7 thesis MSc students have graduated, alongside 12 MSc final projects and 9 research assistants.
- Udi Aharon, API Security
- Aviad Levi, Auction Theory and Machine Learning
- Omri Arbiv, Medical Data Science
- Natan Dillbary, Time and Spectral Time Series Analysis
- Bar Vinder, Medical Data Analysis
- Zohar Shimhon, Sampling-based Classification
- Tanya Zhelezni, Chatbots for Funnel Optimization
- Ohad Nir, Learning Approaches for the Chasing Repairman Problem
- Eden Shmuel, Using Machine Learning to Personalize Nutritional Treatment
- Rafael Ben Lulu, API Security
- Shaked Michael, Access-based Encrypted Traffic Classification
- Dvir Opatowsky, QoE Prediction for FPS Online Games
- Ihab Hino, Dynamic Multimodal Attribute Completion
- Amit Stein, Multimodal Recommendation for User and Item Cold Start
- Eylon Yehiel, Multi-Modal Learning for Product Attributes
- Liav Cohen, Few-shot ANN for Low-latency Classification in O-RAN Networks
- Stas Susha, QUIC-based Quantum Traffic Classification
- Eylon Katan, Anomaly Detection on PQC
- Idan Buzaglo, Early Detection of Myopia
- Yarin Amar, Real time cognitive feedback system in design education
Encrypted traffic 5Medical AI 4E-commerce and multimodal AI 4Adversarial AI 4Incentive design 1Design education 1
We are looking for PhD and MSc students who want to build models that survive contact with the real world.
Topics we are recruiting for now:
- Adaptive encrypted-traffic AI Retrieval, out-of-distribution detection, QUIC and post-quantum traffic. Email about this topic: Adaptive encrypted-traffic AI
- Adversarial and trustworthy ML Robust models for security, healthcare and multimodal systems. Email about this topic: Adversarial and trustworthy ML
- Clinical machine learning Decision support, personalization, privacy and responsible validation. Email about this topic: Clinical machine learning
- Multimodal retrieval Vision-language representations for products, places and recommendations. Email about this topic: Multimodal retrieval
- AI and mechanism design Incentives, marketplaces, teams and resource allocation. Email about this topic: AI and mechanism design
Funding and partners
Competitive grants as PI or co-PI.
₪3.3 million awarded in total
PI Co-PI
Encrypted trafficMedical AIAdversarial AIOther
-
Multi-layered Classification of PQC-Encrypted Traffic
-
Clinical Progression of Dystrophinopathies: Exploring the Role of Autonomic Nervous System Activity and Machine Learning Models
-
Encrypted Networks Traffic Monitoring
-
APK Malware Detection and Analysis
-
Quality of Service Monitoring
-
Violence Prediction in Public Events
Funded by
Partners
- NVIDIAUniversity Ambassador and Deep Learning Institute instructor
- CrySyS Lab, Budapest University of Technology and EconomicsFederated learning security
- KIOS Centre of Excellence, University of CyprusAngelos Marnerides; post-quantum and encrypted traffic
- Transformative Learning Places (TLP) Lab, Ariel UniversityHadas Sopher; design education
- RADCOMCellular traffic data
- Flash Networks
Background
- 2010 BSc, Computer Engineering
- 2012 MSc, Computer Engineering
- 2016 PhD, Computer SciencePostdoc at Vanderbilt
- 2018 Lecturer at ArielDirects the Data Science Research Center
- 2021 Heads the Data Science track
- 2022 Senior Lecturer, tenuredChairs the Faculty Review Board
- 2025 Associate ProfessorNVIDIA University Ambassador
PhD at Bar-Ilan with Prof. David Sarne and Prof. Avinatan Hassidim on mechanisms for online markets; postdoc at Vanderbilt with Prof. Yevgeniy Vorobeychik on adversarial machine learning and crowdsourcing for medical data.