PSST

The unabated rise of smart speech technology, much of which is operated by entities outside of the European Union, has revolutionised consumer access to information and applications using the most convenient interface of all - the human voice. Recently, however, we have come to understand that smart speakers, smartwatches and other voice-operated devices may expose their users to privacy risks. Service operators and hackers alike can access not only spoken words but may also exploit sensitive non-lexical information contained within speech signals (such as health status) for nefarious purposes without the users’ knowledge or consent. However, innovative privacy-preserving smart speech technologies may mitigate these risks and help to create sustainable business models that are well aligned with EU regulations such as the GDPR and the forthcoming EU Artificial Intelligence (AI) Act. Therefore, the PSST joint doctoral training network will train a new cohort of researchers to develop these urgently-needed technologies using cutting-edge deep learning methods. PSST will focus on the suppression of speaker traits in speech signals, on enhancing usability in interactions between smart devices and cloud services, and on novel evaluation schemes that consider the most relevant threats. The PSST consortium is composed of world-leading research labs in speech processing, privacy and usability, and a larger number of innovative industrial partners. The joint academic-industrial partnership will establish a high-impact training programme that provides doctoral researchers with expertise in speech processing, machine learning and comprehensive knowledge of privacy threats and privacy-enhancing technologies. PSST partners and researchers will jointly create not only a unique, user-centred view of smart speech technologies but will also promote innovative business models for responsible and sustainable privacy-aware technologies in the EU.

2025

2030