LPRM at the 35th UFES Scientific Initiation Journey (JIC 2025)
The Laboratory of Research in Networks and Multimedia (LPRM) participated in the 35th UFES Scientific Initiation Journey (JIC 2025), an annual event organized by the UFES Office of Research and Graduate Studies (PRPPG) that showcases the outcomes of undergraduate research projects across all areas of knowledge.
Student Johann Jakob S. Bastos, supervised by professors Rodolfo da Silva Villaça and Vinícius F. S. Mota, presented the project “MininetFed: a tool for integration between real and emulated federated learning environments.”
The work continues the development of MininetFed, a framework for emulating heterogeneous federated learning scenarios.
In this new phase, Johann extended MininetFed to support IoT networks based on the IEEE 802.15.4 standard, using the RPL protocol and Containernet, in addition to implementing complementary energy consumption models and a graphical user interface (GUI) for experiment visualization.
The tool — awarded 3rd Best Tool at SBRC 2024 and recently presented at ACM SIGCOMM 2025 (Coimbra, Portugal) — reinforces LPRM’s contribution to bridging academic research and practical applications in intelligent networking and distributed learning.
Student João Batista, also a member of LPRM, presented the work “Detection of Malicious Clients in Label-Flipping Attack Scenarios with Compressed Models.”
The project proposes a new algorithm for detecting adversarial clients in federated learning under label-flipping attacks.
The method introduces an innovative strategy: local models are sent in compressed form (count-min sketches) along with their uncompressed last layer, which is used to identify outliers via the Modified Z-Score statistical metric, robust to asymmetric distributions.
The results demonstrated strong performance — achieving up to a 10% accuracy improvement in scenarios with up to 30% malicious clients, maintaining accuracy close to that of a benign environment.
These works highlight LPRM’s commitment to fostering new researchers and advancing the state of the art in federated learning, IoT networking, and security in distributed systems.
Congratulations to Johann, João, and their advisors for their excellent work and valuable contributions presented at JIC 2025!
