TRAICT – Trusted Resource Aware ICT

The Fraunhofer-wide project TRA-ICT (Trusted Resource Aware ICT) addresses current issues in the fields of Green ICT and Trusted Electronics. In the project, Fraunhofer IDMT provides software components for energy-efficient, trustworthy acoustic monitoring and thus aims to contribute to the development of suitable hardware for use in mobile and 5G scenarios.

An area of application for energy-efficient and resource-saving AI is acoustic monitoring for traffic surveillance, in which aspects of security and technical data protection also play an important role. Fraunhofer IDMT develops AI-based algorithms for the robust detection and classification of acoustic events. The focus here is on the most energy-efficient and robust audio analysis possible for use on mobile devices on which the classification is to be performed directly. Compressed AI models are used, which provide the best possible performance with less computing power and thus can achieve high accuracy in the analysis.

In addition to energy efficiency, there are also various security and data protection requirements to be met. Therefore, the project develops methods to authenticate, confidentially store and transfer recorded audio data, AI models and annotations. Especially data protection plays an important role in acoustic traffic monitoring, so methods for speech filtering or manipulation are developed. In addition, a novel method for the detection of replay attacks based on microphone classification, among others, is being tested. In the future, numerous applications of acoustic monitoring, such as machine, process or construction site monitoring, protection of critical infrastructure, biodiversity measurement or recording devices for security applications will be realized.

Responsibilities of Fraunhofer IDMT

  • Automatic and robust classification of traffic noise using compressed AI models for mobile use
  • Development of methods for authentication and secure processing of audio data, AI models and annotations
  • Recognition of replay attacks based on microphone classification

Partner

Fraunhofer Institutes IIS, AISEC, EMFT, ENAS, FOKUS, IAF, IAIS, IESE, IGD, IMS, IMWS, IOSB, IPMS, ISIT, ITWM, IZFP, IZM

Sponsors

Fraunhofer internal funding program

Duration

July 2020 – December 2020