[Startup Bharat] This platform founded by IIT alumni and an AI scientist is detecting deepfakes

Deepfake videos are emerging to be the most dangerous form of crime. Deepfakes involve manipulating videos to show someone saying or doing something they never said or did. According to a report by University College London (UCL), fake content would be difficult to detect and stop, and that it could have a variety of aims – from discrediting a public figure to extracting funds by impersonating a couple’s son or daughter in a video call.

The rising trend in deepfakes has raised concerns regarding the “destructive potential”, and the incidents have opened the global debate over fake posts and deepfakes, for which Kroop AI, a Gandhinagar-based startup, has developed a solution.

Started in February 2021 by Jyoti Joshi, an AI scientist; and IIT alumni Milan Chaudhari and Sarthak Gupta, Kroop AI is a deployable AI-enabled platform for businesses and individuals to identify deepfakes across audio, video, or image data. The platform is deep learning-based, which detects and analyses deepfakes in detail across different platforms and information mediums.

“There have already been several reports coming in stating the misuse of deepfake technology to commit millions of dollars of fraud,” says 33-year-old Jyoti.

The Gandhinagar-based startup, with a team of ten people, wants to become a global affordable tool for people and businesses to detect fake content at affordable prices. The six-month-old startup is in its pre-revenue stage, but is targeting the banking and finance sector, and is already working with an entertainment company in India.

From research to startup

In 2019, IIT Ropar friends Sarthak Gupta and Milan Chaudhari were doing multiple tech projects and were hoping to find a unique idea to start up. One of their key interests was artificial intelligence, thanks to their mentor and professor Abhinav Dhall.

While working as an AI researcher in Australia at Monash University, Jyoti was specialising on tools to fight threats of deepfakes and worked with Abhinav for her research work. Towards the end of 2020, Sarthak Gupta and Milan Chaudhari got together to discuss the advent of synthetic data generation technology and the potential damage deepfake technology could cause.

The connecting dot between the team was Abhinav, who had worked with all three of them at different stages. He connected Jyoti with Sarthak and Milan to brainstorm on the technology to tackle deepfakes. This laid the foundation of Kroop AI.

ALSO READ

[Startup Bharat] How Bhubaneswar-based Impact Health is looking to disrupt end-to-end healthcare with tech and logistics

Recalling early days, Jyoti says that to challenge and test, the plausibility of misuse of deepfakes with free and easily accessible deepfake creation software, the trio experimented using ultra high quality live deepfake during a video conferencing call amongst themselves. Thus, they were convinced that as synthetic data generation grows rapidly, it is going to widen the gap in the market, creating a huge demand for solutions to check for the validity and authenticity of the content.

“Deepfakes not only pose serious threats to individuals, but also endanger large corporations and businesses across different sectors,” says Jyoti.

She also emphasises that high-quality deepfakes are capable of altering the history as perceived as it may aid propagandists and conspiracy theorists to support and highlight an alternative narrative.

“Not only this, deepfakes hold potential to deceive currently deployed biometric systems based on face or speaker authentication,” Jyoti adds.

With a vision to project reality by uncovering deception, the founders got together to build a deepfake detection platform that marked the inception of a joint venture, which later formalised as Kroop AI.

The three bootstrapped and started Kroop AI formally at the start of 2021 with Rs 5 lakh investment.

Competition and challenges

The startup said it faces competition from a few companies based out of the US and Europe like Sentinel, Sensity, and Quantum Integrity, but Jyoti says that Kroop AI is a unique platform, which offers detection for all modalities that are audio-only, video only, audio-visual, and image.

“The solution is capable of detecting different ways in which deepfakes could have been administered such as facial expression transformation, face swap or facial and audio re-enactment. The solution also provides analytics specifying localisation in space and time such that users can make an informed decision on the authenticity of the content,” she says.

She also adds that the platform does not only detect fake media but can also ethically generate synthetic data for B2B use cases.

However, Jyoti says, there are more challenges than the competition. The real challenge is to bring awareness to masses about deepfakes and the potential harm they could cause.

“The more the first-hand consumers of manipulated content are aware, the easier it would become to avoid the frauds and misuse of deepfakes. Kroop AI has also taken the initiative of creating vlogs to educate on subtle cues one could look for while identifying fake content,” she adds.

Revenue model, funding, and future plans

Kroop AI provides deepfake detection as a service. As of now, the startup is targeting BFSI (securing video KYC) and cybersecurity. The pricing is based on the quantum and duration of audio-visual data analysed using its API. Jyoti says that the startup has already bagged one client, and is in talks with banking companies in India.

For the next 12 months, the main focus of the startup is on the B2B and B2B2C sector. “There already is keen interest to uptake our deepfake detection technology in Indian BFSI domain by integrating this essential security measure to defend against any fraudulent attempts,” says Jyoti.

Going forward, the company wants to raise another round of funding next year to expand its global footprint, and work towards the ethical generation of synthetic data. “The high-quality deepfakes will help the unique Kroop AI signature embedded in it which won’t be observable but can be accessed through our API,” says Jyoti.

The company has a total investment of $180,000 from 100X.VC and angel investors. It is currently using the funds to expand its research and development team, developers, and supporting operations of the company. “It will also help us to take our product into global markets,” says Jyoti.