Research Project
Privacy Smart Speech Technology
The rise of smart speech technology has revolutionised consumer access to information and applications. However, privacy risks, including potential breaches by service operators and hackers accessing spoken words and sensitive non-lexical information, are a concern. Privacy-preserving smart speech technologies could help mitigate these risks. With the support of the Marie Skłodowska-Curie Actions programme, the PSST project aims to develop urgently needed technologies using deep learning.
PSST!
-
Europe
1. Project Overview
Smart speech technologies are increasingly used in speakers, watches, mobile devices, connected services, and voice-enabled applications. While these systems make interaction more natural and convenient, they also introduce privacy risks that are often invisible to users. Voice data can contain sensitive information beyond spoken words, including speaker identity, health-related indicators, emotional state, demographic traits, and other personal characteristics. This creates challenges for service providers, regulators, researchers, and technology developers working to ensure that speech systems remain useful while protecting user privacy. PSST is a Horizon Europe Marie Skłodowska-Curie doctoral training network focused on developing privacy-preserving smart speech technologies. The project trains doctoral researchers in speech processing, machine learning, privacy assessment, and usability, while exploring new methods for anonymization, secure device-cloud interaction, and privacy threat evaluation. The project is organized around three research teams — Defence, Attack, and Utility — creating an iterative research framework where privacy-preserving methods are developed, challenged, and evaluated for practical use.
2. VoiceInteraction's Contribution
VoiceInteraction contributes industrial expertise in speech processing, deployable speech technologies, and operational AI systems. Within PSST, VoiceInteraction supports the connection between academic research and practical speech technology applications, helping doctoral researchers understand how privacy-preserving methods can be evaluated in real-world speech environments. The project aligns closely with VoiceInteraction’s work in trustworthy AI, secure deployment, multilingual speech processing, and privacy-aware speech systems, especially for organizations that require control over sensitive spoken data.
3. Key Capabilities
Privacy-preserving speech processing
Speech anonymization and identity protection
Smart speech technology evaluation
Privacy threat assessment
Human-centric speech AI
Device-cloud speech interaction
Deep learning for speech privacy
Trustworthy and responsible AI
GDPR and AI Act-aligned technology development
Academic-industrial research collaboration
4. Outcomes & Impact
PSST contributes to the development of future speech technologies that are more privacy-aware, transparent, and aligned with European values around trustworthy AI. For VoiceInteraction, the project strengthens research capabilities in privacy-preserving speech processing, evaluation methodologies, and responsible AI deployment. The knowledge developed through PSST supports future applications where speech technologies must balance usability, performance, and protection of sensitive user information. The project also reinforces VoiceInteraction’s role as an industrial partner capable of supporting doctoral research, technology validation, and the transition from academic research into practical speech technology applications.
5. Partners & Consortium
PSST is a Horizon Europe Marie Skłodowska-Curie doctoral training network involving 7 European universities and 11 industrial partners, bringing together expertise in speech processing, machine learning, privacy, usability, smart speech technologies, and responsible AI.
6. Project Information
Project Code
10.3030/101168193
Funding Programme
MSCA
Intervention Region
Europe
Beneficiary Entity
VoiceInteraction - No EU contribution
Official Project Site →
7. Funding Information
HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA)
← Back to all projects
Operational speech workflows require different approaches
Discuss transcription, monitoring, accessibility, or conversational analysis requirements with the VoiceInteraction team.