Cyberagentur launches two projects on next-generation audio forensics
With its Audio Forensics (AuFo) research programme, the Agentur für Innovation in der Cybersicherheit GmbH (Cyberagentur) is financing two complementary research approaches that aim to extract previously hidden meta-information from low-quality audio data. The aim is to enable investigative authorities to act faster, more precisely and more reliably in time-critical situations.
As part of the Audio Forensics (AuFo) research programme, the Cyberagentur has hired two project partners who are breaking new ground in the forensic analysis of audio data with different, complementary approaches. AuFo addresses a field that has received little research to date: the systematic evaluation of room acoustics and technical metadata from low-quality audio recordings, in the case of emergency calls, missing persons cases or secretly made audio recordings. While previous research has focused primarily on speech content and speaker identification, AuFo focuses on reconstructing the recording environment.
The programme aims to draw conclusions about the characteristics of the recording location from acoustic signals – such as room size, geometry, material properties or the distinction between indoor and outdoor spaces. Ideally, this should also enable regional delimitation. Additionally, the project is investigating whether the distinctive features of the recording device, such as a smartphone microphone, can be identified. Even initial reliable investigative clues can be of considerable value to law enforcement in time-critical scenarios.
The AnFAng project – Applied Research for Audio Forensic Investigations – is led by marinom GmbH as the contractor with the University of Bremen (Institute for Underwater Acoustics, Sonar Technology and Signal Theory) as the subcontractor. Their approach is to develop a system for the near real-time analysis of audio signals. With the help of the developed system, investigative authorities should be able to derive acoustic clues and metadata in order to obtain information about the caller’s location beyond the voice content itself. Key application scenarios include emergency calls, supplemented by issues of counter-terrorism and external security. The robotic agent RAFA (Robotic Agent for Field Acoustics) generates a high-resolution reference data set through detailed room measurements. The signal condition is analysed before applying common filters such as noise reduction in order to identify reliable room acoustic characterisations. These can be used to make statements about the room’s geometry and nature even under poor conditions and that will also stand up to subsequent legal scrutiny.
The project SAFIRA – Sound & Audio Forensic Investigation and Research Activity is led by Binaurics Audio GmbH as the contractor. The subcontractors are Ferncast GmbH and the Institute for Communication Systems at RWTH Aachen University. Dr. Patrick Naylor from Imperial College London is contributing international expertise as an associated partner. SAFIRA aims to develop robust forensic methods for characterising and identifying recording locations even under adverse conditions, such as highly noisy, compressed or only partially referenceable audio data. The core of the approach is the creation of a dynamically expandable “acoustic fingerprint” based on room acoustic parameters. This will be used to map temporal sequences and, in the future, enable geotagging of specific locations. This is based on an audio forensics database containing recordings from various locations and sources, including messengers and social media services. A hybrid approach combining classic signal processing and machine learning based strategies is being investigated, supplemented by simulation tools for auditory and automated result verification. Reliability and confidence measures as well as the option for manual intervention are an integral part of meeting the requirements of real forensic practice.
“With AuFo, we are deliberately breaking new scientific ground. It is currently unclear whether and to what extent recording locations can be reconstructed from low-quality audio data. But that is precisely where this programme’s disruptive potential lies,” emphasises Olivia Gräupner, project manager and research specialist in the Cyberagentur’s Secure Society department. “The two projects pursue different methodological approaches and address key risks and opportunities at the same time. Our aim is to create a solid foundation that will enable investigative authorities to act more quickly in highly sensitive situations – even if success is not yet guaranteed at this stage.”
From a scientific and technical perspective, AuFo is classified as high-risk research. The Cyberagentur is therefore financing disruptive risk research in order to develop new security-related capabilities that go beyond current technology and strengthen the long-term capacity of state actors to act in digital and hybrid spaces.
Further information: