Students win prize for deepfake detection tool for banks

Winning a national hackathon

Three students from the Thapar Institute of Engineering and Technology recently secured first place at the Rakshak-2026 National AI Cybersecurity Hackathon. The competition, which took place at IIT Delhi, saw more than 4,700 registrations from across the country. Out of those entries, only 30 teams eventually presented their solutions to the judges.

The winning team, which called themselves 'Mama Mia,' received a cash award of Rs 2 lakh for their work. The group consisted of Gनी Singh, Yash Bishnoi, and Keshav Juneja. They were tasked with creating a prototype that could address the growing threat of digital identity theft in the financial sector.

The students were required to build their solution during a strict 40-hour development period. Despite the short timeframe, the team managed to create a functional face-detection website. This rapid development showcased their ability to turn a concept into a working prototype under high-pressure conditions.

How the detection works

The core of the students' project is a system designed to identify deepfakes, which are videos where a person's likeness is digitally altered or replaced. The tool is specifically aimed at protecting video-based KYC, the process banks use to verify a customer's identity. If a person on a video call is using a fake face, the system is meant to flag the activity immediately.

To determine if a video is genuine, the software examines signals across four distinct categories of checks. These categories include the specific details of facial features and the way a person moves their body. It also looks for technical indicators that are linked to the camera sensor itself to find inconsistencies.

Once the system completes its analysis, it provides an alert to the banking agent handling the verification. This allows the agent to stop the process or perform a more thorough review if the software detects suspicious activity. The goal is to provide a layer of security that can catch manipulations that might fool a human observer.

Banking and business use

The students designed the architecture so that it can operate through both a web browser and a mobile application. This flexibility ensures that different types of banking interfaces can utilize the technology. It is primarily intended as a business-to-business solution, meaning it is built for banks to use rather than for individual consumers to own.

The team plans to integrate this technology into existing KYC procedures already used by financial institutions. They also intend to keep access to the tool free for the end customers. This means that while the bank pays for the underlying technology, the people being verified do not have to pay an extra fee to use it.

During the development of the prototype, the students focused on making the tool compatible with current requirements. They specifically considered the guidelines set by the Reserve Bank of India. By aligning the technology with these regulations, the team aimed to make it easier for established banks to adopt the system into their current workflows.

The path forward

While the prototype was successful in the competition, it has not yet been deployed by any specific bank. The report did not include information regarding any formal regulatory approvals or independent testing of the system outside of the hackathon environment. The next phase for the team involves further technology development to refine the prototype.

To help with this next stage, Amazon is expected to provide selected hackathon teams with Amazon Web Services credits. These credits will help the students continue to build and improve their technologies. This support is part of a broader effort to encourage the development of tools that combat cyber threats.

The hackathon also highlighted the importance of public awareness regarding digital fraud. Amazon India's vice-president for legal, Rakesh Bakshi, noted that companies should focus on educating the public about these risks. The event served as a platform for students to demonstrate how AI can be used to defend against unlawful activities like identity cloning.