News
Salma is awarded the Early Career Teaching Excellence award by the UCI SSoE!
Salma is awarded Professor of the Year (2022-2023)
Engineering Student Council (esc) in the Henry Samueli School of Engineering has awarded Prof. Elmalaki the Professor of the Year for the Electrical Engineering and Computer Science Department (EECS). This award is a student-nominated, voted, and organized award.
New Paper at CPS-IoT week 2023 (IoTDI’23). Congrats Mojtaba!
adaParl : Adaptive Privacy-Aware Reinforcement Learning for
Sequential-Decision Making Human-in-the-Loop Systems
Our new work on adaptive privacy-aware RL that captures human variability in sequential-decision making systems has been accepted to appear in IoTDI’23. Congratulations, Mojtaba and Stelios!
New Accepted paper at the 32nd USENIX Security. Congrats Hieu!
A new paper by Hieu Le, “AutoFR: Automated Filter Rule Generation for Adblocking” has been accepted in the 32nd USENIX Security. Congratulations Hieu Le for your tremendous effort and commitment.
This paper provides the very first full-stack system for automatic filter rules generation for Adblocking that is adaptive and tunable using a novel Reinforcement Learning modeling. Unique to this paper is the modeling of the user’s preference in the tradeoff between blocking ads and avoiding breakage. We proposed a multi-arm bandit model to generate filter rules while controlling this trade-off in an automatic and systematic approach. I believe this paper will have a huge impact on the curation of the next-generation filter rules. Stay tuned for the full article!
This paper is a joint collaboration with Athina Markopoulou and Zubair Shafiq. Congratulations!
Invited Talk: Human Factors in Intelligent Vehicles (HFIV’22) – Workshop Series at IEEE IV/ITSS Conferences
Salma is invited to give a talk at the Workshop on Human Factors in Intelligent Vehicles at Intelligent Vehicle Symposium based on her single-authored paper at the IV’22.
New Single-Authored Paper: “MAConAuto: Framework for Mobile-Assisted Human-in-the-Loop Automotive System” has been accepted by the IEEE Intelligent Transportation Systems Society to appear in the 33rd IEEE Intelligent Vehicles Symposium (IV).
In this paper, we propose MAConAuto, a general framework to design and develop automotive applications as human-centric applications. MAConAuto aims at improving the human experience by integrating the perception and reactions of the human into the interventions made by the automotive applications through monitoring the human response. Modeling the human in a way that captures the change in perception and reaction is an open challenging research question. Borrowing up from the psychology literature, the behavior of the changes in the human decision historically was modeled through the expected utility theorem (EUT) which is based on an axiomatic framework defined as completeness, transitivity, independence, and continuity. Human is said to be rational if these four axioms hold. However, the models that rely on the EUT have shown that these axioms are unrealistic and that human decisions tend to deviate from the axioms of the EUT. In this paper, we dissect these axioms to design a human model for human-in-the-loop automotive applications using Multisample Reinforcement Learning.
Talk by Hieu Le in AdblockerDevSummit 2021
New talk by Hieu Le in the 2021 AdblockerDevSummit on our exciting work for automatic generating and testing of filter rules. This work in collaboration with Athina Markopoulou (UCI) and Zubair Shafiq (UCD)
New Paper: Future of Smart Classroom in the Era of Wearable Neurotechnology.
Check out our new paper on the challenges and approaches for the future of smart classrooms using wearable neurotechnology.
Mojtaba Taherisadr, Berken Utku Demirel, Mohammad Abdullah Al Faruque, Salma Elmalaki, “Future of Smart Classroom in the Era of Wearable Neurotechnology“, October 2021, arXiv:2110.11475