A man presents scientific data on a screen at a conference podium.

Lucky Maduabuchi Onyekwelu-Udoka, a CSAFE doctoral researcher, presented his research at the 2025 American Academy of Forensic Sciences annual scientific conference.

Posted December 3, 2025


Digital recordings are now a common feature of criminal investigations, emergency calls, civil disputes and online misinformation. However, as audio becomes easier to manipulate, forensic examiners need stronger scientific tools to determine what is genuine, who is speaking and whether a recording has been altered.
At the Center for Statistics and Applications in Forensic Evidence (CSAFE), Lucky Maduabuchi Onyekwelu-Udoka is helping to lead this effort. He is a doctoral researcher at Iowa State University, studying under CSAFE researcher Yong Guan, a professor of electrical and computer engineering.

Onyekwelu-Udoka’s work focuses on three rapidly emerging challenges in audio forensics: background noise interpretation, speaker identification in multi-speaker recordings and deep-fake detection.

Understanding Background Noise to Detect Tampering

Background sounds, such as dog barks, sirens, traffic, machinery or footsteps, can reveal important clues about when and where a recording was made. Onyekwelu-Udoka’s research uses AI and statistical modeling to interpret these environmental traces and identify whether different parts of a recording match or if segments may have been inserted, removed or re-recorded. This work strengthens the scientific basis for determining audio authenticity.

Who Is Speaking, and How Many People Are There?

Many forensic recordings include multiple speakers talking to each other. To help examiners untangle these conversations, Onyekwelu-Udoka develops advanced methods for speaker diarization (“who spoke when”) and speaker counting. These tools can detect additional or missing voices and clarify complex recordings from interviews, emergency calls or surveillance footage.

Combating Deepfakes and Synthetic Speech

As AI-generated voices become more realistic, distinguishing real speech from cloned or machine-produced audio is critical. His research builds detection models that identify subtle artifacts left behind by text-to-speech and voice-conversion systems, even when background noise or poor audio quality is present. This supports national efforts to maintain the reliability of digital evidence.

Why This Work Matters

Onyekwelu-Udoka explained that he was motivated by seeing how easily audio can be manipulated and how such manipulation can influence legal outcomes. “Ensuring the scientific integrity of audio evidence is essential for fairness in the justice system,” he said.

He views CSAFE as the ideal environment to advance this mission, noting that its emphasis on statistics, transparency and interdisciplinary collaboration helped shape his research direction. His ongoing work contributes to CSAFE’s broader goals of improving forensic accuracy, strengthening courtroom testimony and supporting practitioners across the country.

Looking Ahead

The next phase of his research focuses on expanding validation studies, improving the uncertainty of communication for courtroom testimony and developing practical tools that forensic laboratories can adopt.

Through this work, Onyekwelu-Udoka is helping to build the next generation of scientifically grounded methods for analyzing digital audio, a field that is growing in importance as technology evolves.

At the 2025 American Academy of Forensic Sciences annual scientific conference, Onyekwelu-Udoka presented “Environmental Signature Extraction for Forensic Analysis of Audio Recordings Using AI/LLM Models.” View the presentation slides on the CSAFE website.

From the Blog

Journal article titled

Two Journals for Sharing Your Forensic Science and Statistics Research

If you’re looking for a journal to showcase your latest research findings, consider exploring these …

NIST sign outside, surrounded by flowers and trees, on a sunny day.

NIST Report Highlights Key Challenges in Forensic Science and Provides Strategies for Addressing Them

In September, the National Institute of Standards and Technology (NIST) released a new report titled …

Recent Posts

ICFIS 2026 logo features a clock tower and the Forensic Inference and Statistics name.

Keynote Speakers Announced for ICFIS 2026

August 10, 2026

The Scientific Committee for the 12th International Conference on Forensic Inference and Statistics is pleased to announce the keynote speakers …

2026 Wisconsin NAFSB Conference logo on a dark blue background.

CSAFE Researchers will be Panel Speakers at NAFSB 2026 Conference

August 7, 2026

The annual National Association of Forensic Science Boards (NAFSB) Conference will be Oct. 13-16 at the Milwaukee Crime Lab Bureau …

ICFIS 2026 logo features a clock tower and the Forensic Inference and Statistics name.

ICFIS 2026 Moves Online, Abstract Deadline Extended to Aug. 17

July 27, 2026

The 12th International Conference on Forensic Inference and Statistics (ICFIS) will now be an online-only event, allowing participants from around the world …