Photo Courtesy of Qi Hu
SACRAMENTO, Calif., Aug. 06, 2026 (GLOBE NEWSWIRE) — As the Federal Bureau of Investigation reports that Americans lost more than $16.6 billion to digital financial scams in 2024, a 370% increase over five years, independent AI researcher and computer science graduate Qi Hu has had two studies accepted at the 2026 International Conference on Intelligent Computing (ICIC 2026) in Toronto, Canada, each targeting a technical gap that allows financial fraud to flourish.
Hu, who specializes in financial intelligence and computer vision, will present two frameworks — ProtoHGC and RoR-CV — at the conference on July 22–26, 2026. Both address problems U.S. regulators have flagged as critical weaknesses in the nation’s financial infrastructure: detecting sophisticated fraud, and making AI-driven financial decisions explainable and auditable. The papers were selected through ICIC’s peer review process and will be published by Springer in its Lecture Notes in Computer Science (LNCS) series.
Despite the scale of the crisis, with over 76% of U.S. organizations experiencing attempted or actual payment fraud in 2025, and only 17% currently using AI to combat it, financial institutions continue to struggle with detecting sophisticated fraud patterns hidden within complex transaction networks.
ProtoHGC maps out the web of connections on a digital payment platform — how accounts, transactions, and timing relate to one another — to spot the patterns that signal fraud. Because it learns what suspicious activity looks like from the structure of the data itself, it can flag fraud even when there are few confirmed examples to learn from. In benchmark testing, ProtoHGC detected fraud more accurately than the leading alternative methods across every measure evaluated, according to Hu.
U.S. financial regulators have increasingly emphasized explainability and auditability as prerequisites for responsible AI deployment in financial services. Conventional systems often produce accurate outputs without traceable rationales, a critical liability in regulated environments.
RoR-CV brings together three things a human analyst would use — charts and financial reports, an understanding of cause and effect, and awareness of shifting market conditions — into a single tool that recommends investments and explains its reasoning. It reads the figures in charts and reports, traces how broad economic events ripple down to affect specific companies and assets, and adjusts how much weight it gives visual data as markets grow more volatile. According to Hu, in testing, RoR-CV made more accurate and better-explained recommendations than existing tools, including during downturns and turbulent markets.
These acceptances build on Hu’s earlier peer-reviewed work, including a method for long-sequence video object segmentation published in IEEE Signal Processing Letters (2026), an AI risk-control system for online lending presented at an ACM conference (2025), and a financial recommendation tool that outperformed a leading industry model by roughly 10–12% on standard accuracy measures. Hu’s published research has been cited 44 times, and she serves as a peer reviewer for journals including IEEE Access, Pattern Recognition Letters (Elsevier), and Informatica.
“The combination of graph-based fraud detection, explainable multimodal recommendation, and AI-driven risk management represents a comprehensive approach to one of the most pressing challenges in U.S. financial technology,” said Qi Hu.
About ICIC 2026
The International Conference on Intelligent Computing (ICIC) is an annual gathering of researchers in artificial intelligence and related fields, now in its 22nd edition. The 2026 conference takes place July 22–26 in Toronto, Canada, bringing together academics and practitioners to present new work across machine learning, computer vision, and applied AI.
Papers accepted at ICIC are published by Springer in its Lecture Notes in Computer Science (LNCS) and Lecture Notes in Artificial Intelligence (LNAI) series, long-established references in the field.
Contact Information:
Contact Name: Audrey Smith
Company: Sparker Lab LLC
Company website: sparkerlab.com
Contact Email Address: [email protected]
Sacramento, CA 95814, USA
A photo accompanying this announcement is available at https://www.globenewswire.com/NewsRoom/AttachmentNg/ac43b0d0-5413-46a4-9be1-6111e6ef64bd
