Talk 1: Effective but Unrealistic: Addressing NIDS Challenges in the Real World
Abstract: Network Intrusion Detection Systems (NIDS) are crucial mechanisms of cybersecurity and are widely used to protect network organizations. Recent literature has increasingly highlighted the vulnerability of NIDS to adversarial evasion attacks (AEAs) and proposed countermeasures. However, existing literature on AEA methods, while effective, often lacks realism. These approaches predominantly originate from the image domain, where application scenarios significantly differ from those in network settings. This discrepancy hinders the direct transposition of lab-conceived attacks to real-world environments. This talk sheds light on the challenges encountered in adapting AEA strategies from the image to the network domain. Furthermore, we present two methods that are realistically viable in generating adversarial evasion attacks for the NIDS domain, offering practical insights into the development of robust defense mechanisms.
Talk 2: A soft sensor to assess the energy performance of laundry washing machines
Abstract: In the EU, over 15 million washing machines are purchased annually, accounting for nearly 5% of domestic electricity consumption. This study presents a low-cost soft sensor method for assessing the real-world energy performance of domestic washing machines based on resource efficiency. The method utilizes soft sensor techniques to estimate load mass and performance indicators from power and water consumption data collected by an IoT-enabled sensor module. The indicators considered are energy and water consumption per kilogram of laundry load. These indicators are combined into a single metric for comprehensive energy performance evaluation. The method is tested on six washing machine models, focusing on the standard ‘Cotton’ 40°C washing cycle and varying the laundry load mass. The proposed method can be applied in real-world scenarios to assess washing machine performance and recommend optimal settings to maximize energy savings.
Allan Espindola is pursuing his Ph.D. in Computer Science under a joint doctoral supervision program. He is collaboratively guided by Professors António Casimiro and Pedro Ferreira from the University of Lisbon, Faculty of Sciences, as well as Professors Altair Santin and Eduardo Viegas from Pontifícia Universidade Católica do Paraná, Escola Politécnica. Allan’s research is centered on enhancing the robustness of network intrusion detection systems. He specifically focuses on mitigating adversarial evasion attacks by leveraging the concept of diversity in system design and defense strategies.
Zygimantas Jasiunas is currently pursuing a Ph.D. at the Department of Informatics within the Faculty of Sciences at the University of Lisbon. His research focuses on evaluating and optimizing appliances through the utilization of embedded systems and Internet of Things (IoT) technologies, coupled with machine learning methodologies. Zygimantas is conducting his research under the guidance of Professors Pedro M. Ferreira and Jose Cecilio.