Tanveer Khan from Tampere University (TUNI) represented the HARPOCRATES project at the 19th EAI International Conference on Security and Privacy in Communication Networks (SecureComm 2023), held on 19–21 October 2023 in Hong Kong, Japan.
SecureComm is a well-established international venue focusing on the latest research in secure communications, privacy-preserving systems, and network resilience. It brings together researchers, practitioners, and industry experts to discuss advances in both theoretical foundations and applied security technologies.
Research Contribution
At the conference, Tanveer Khan presented the paper:
“Split Without a Leak: Reducing Privacy Leakage in Split Learning.”
The paper addresses privacy concerns in split learning, a machine learning paradigm where training is divided between client and server to reduce computational burden. While split learning improves efficiency, it introduces risks of privacy leakage, as sensitive information can be inferred from intermediate representations shared between parties.
The research proposes novel techniques to mitigate these risks, ensuring that collaborative training can be performed without compromising sensitive data. This contribution aligns closely with HARPOCRATES’s mission of advancing privacy-preserving machine learning (PPML) and strengthening trust in secure, data-driven applications.
The HARPOCRATES project develops cutting-edge methods for confidentiality, integrity, and resilience in AI systems, with a particular focus on privacy-preserving approaches such as functional encryption, hybrid homomorphic encryption, and secure multiparty computation. Tanveer Khan’s work on split learning contributes directly to these objectives, demonstrating practical ways to reduce vulnerabilities in distributed AI systems.
Further Information
Details on the conference are available on the official website: SecureComm 2023