As artificial intelligence makes it increasingly difficult to distinguish authentic content from fabricated images, videos and audio, Nigerian-born cloud engineer Louis Echefu is working to make deepfake detection faster, more accessible and capable of protecting digital users.

The rapid advancement of artificial intelligence has transformed how people create and consume digital content. But the same technology driving innovation is also making it easier to fabricate convincing images, imitate human voices and manipulate videos.

From financial fraud and identity theft to misinformation, deepfakes are increasingly challenging public confidence in what people see and hear online.

For Echefu, a Cloud Consultant specialising in application development at Amazon Web Services, the challenge goes beyond simply identifying manipulated content.

He says, “At its simplest, it means developing technology that can examine an image, video, or audio recording and determine whether artificial intelligence was used to create or manipulate it. But identifying a fake is only one part of the challenge. The bigger question is how to make that technology work efficiently when millions of digital files are being shared every day.

“That is where cloud computing becomes important. It provides the computing capacity needed to process large amounts of information, update detection models as new threats emerge, and deliver results quickly. My interest is in making these capabilities practical beyond research laboratories, so that financial institutions, media organizations, government agencies, and ordinary users can benefit from them.”

Echefu’s research combines artificial intelligence, data engineering, cybersecurity and cloud computing in an effort to improve the way digitally manipulated content is identified.

His 2024 peer-reviewed study, Evaluating Deepfake Images: An Empirical Evaluation of Select Methods with Data Engineering, published in the Journal of Multimedia Information System*, reported a detection accuracy of 98.97 per cent under the study’s experimental conditions.

He has also developed a publicly accessible tool that allows users to assess whether an image may have been generated or manipulated using artificial intelligence.

Echefu said the growing threat posed by deepfakes was particularly concerning because of its potential to undermine public trust.

“The biggest concern is that they undermine something society depends on: trust. We have reached a point where someone can create a convincing image of a public official, imitate the voice of a business executive, or fabricate a video of an event that never happened.

“The consequences can be serious. A manipulated recording could be used to authorize a fraudulent transaction, damage someone’s reputation, or spread misinformation during an election. Women, children, and other vulnerable individuals can also become targets of maliciously fabricated content.

“What concerns me is the imbalance between how easily these materials can be created and how difficult they can be to verify. Powerful AI tools are becoming widely available, but reliable detection capabilities are not equally accessible. My research is motivated by the need to narrow that gap”, he said.

According to him, the rapid evolution of generative artificial intelligence means deepfake detection cannot be treated as a problem that can be solved permanently with a single model.

“Because deepfake detection is not a problem that can be solved once and forgotten. Artificial intelligence is constantly evolving. As image and video generation systems become more sophisticated, the patterns that detection models previously relied upon can become less useful.

“A model that performs exceptionally well on one dataset may not achieve the same results when confronted with unfamiliar images or newer generation techniques. That is why continuous testing, retraining, and independent validation are essential.

“Cloud technology makes this process more manageable because models can be updated and evaluated without rebuilding an entire system. The goal should not be to develop a detector that performs well today but one that can continue adapting as the technology changes”, he stated.

Echefu said his research also demonstrated the importance of data engineering in developing effective deepfake detection systems.

“One of the most important aspects of the research was understanding how data preparation influences the performance of deepfake detection systems.

“We developed a dataset containing manipulated images of public figures and used it to evaluate different deep-learning approaches. The highest-performing model achieved 98.97 percent accuracy in our experimental evaluation.

“But beyond that figure, the research demonstrated how important data engineering is to building effective detection models. The quality, diversity, and organization of training data can significantly influence a system’s ability to identify manipulation.

“The findings were published in a peer-reviewed journal in 2024, contributing to the broader scientific discussion about improving deepfake detection. Of course, achieving high accuracy under experimental conditions does not mean a model will identify every deepfake in the real world. That is precisely why continued research and validation matter”, he added.

Beyond the research, Echefu has sought to make deepfake detection accessible to ordinary users through a publicly available tool.

“I believe research becomes more meaningful when people can see how it applies to problems they encounter in everyday life.

“Most individuals do not have the technical knowledge or resources to build an artificial intelligence detection system. Yet they are often the people most exposed to suspicious images circulating through social media, messaging platforms, and other digital channels.

“Developing a publicly accessible tool was a way to bring the technology closer to those users. Rather than expecting everyone to understand the technical details of machine learning, the idea is to provide a simpler way to assess potentially manipulated content.

“Such tools are not perfect and should not be treated as absolute proof that something is genuine or fake. However, they can provide an additional layer of verification and encourage people to question suspicious material before accepting or sharing it.”

For Echefu, however, detection is only one part of the response. He believes cloud technology can also help organisations act on suspicious content before it causes widespread damage.

“Detection is useful, but its real value comes from what happens after suspicious content is identified. For example, if a fraudulent video is uploaded to a platform, there should be a mechanism for flagging it, reviewing it, and taking appropriate action before it spreads widely.

“Cloud-based systems can support that process by connecting detection models to automated workflows. Suspicious material can be assessed as it enters a system, prioritized for human review, and documented to support decisions about further action.

“Another important advantage is accessibility. A large technology company may have the resources to develop its own detection infrastructure, but smaller organizations often do not. Cloud services can make these capabilities available without requiring every institution to maintain expensive computing facilities.

“To me, that is one of the most promising aspects of this work: making advanced protective technologies useful to organizations regardless of their size.”

Echefu’s journey into cloud computing and cybersecurity began in Nigeria.

“My background is in engineering. I studied Mechatronics Engineering at Bells University of Technology in Nigeria, which exposed me to several aspects of technology, including hardware, automation, control systems, and software development.

“Over time, I became increasingly interested in the software side, particularly how intelligent systems process information and make decisions. That curiosity eventually led me toward computer science and cybersecurity.

“I pursued a master’s degree at Louisiana State University Shreveport in the United States on a merit scholarship, focusing on computer science and cybersecurity. My academic experience strengthened my research interests and contributed to my work in peer-reviewed publications. I also serve as a reviewer for *IEEE Access*, which allows me to engage with emerging research across technology and engineering.

“Today, my professional work at AWS involves designing and delivering cloud-based application solutions for enterprise and public-sector customers. That experience gives me a practical understanding of how technology must perform outside an academic environment. It has reinforced my belief that effective innovation should be technically sound, scalable, and accessible.”

As artificial intelligence continues to evolve, Echefu believes deepfake detection systems will have to become more sophisticated, faster and more integrated.

“I believe the future will involve systems capable of analyzing images, videos, and audio together, rather than treating them as separate problems. Deepfake threats are becoming more sophisticated, and our defenses must evolve accordingly.

“Speed will also be critical. If harmful content can spread to millions of people within minutes, detecting it several days later may achieve very little. We need systems that can operate much closer to real time while maintaining reliability and appropriate human oversight.

“Cloud infrastructure will be central to making that possible because it allows detection capabilities to be deployed, updated, and expanded across different platforms and institutions.

“Ultimately, the objective is not to prevent people from using artificial intelligence creatively. It is to ensure that technological progress does not come at the expense of security, accountability, and public trust.”

For Echefu, the fight against deepfakes is therefore not simply about developing more accurate algorithms. It is about ensuring that the ability to verify digital information keeps pace with technologies capable of manipulating it.

His research underscores the importance of combining detection models with carefully engineered data, scalable computing infrastructure and practical tools that can reach users beyond major technology companies.

While the 98.97 per cent accuracy reported in his study represents the result of an experimental evaluation, Echefu acknowledges that the rapidly changing nature of artificial intelligence means detection systems must continue to evolve.

As governments, businesses and individuals confront a digital environment where convincing fabricated content can be produced within seconds, his focus remains on making verification a more dependable and accessible part of everyday digital life.

“The goal is not simply to identify one fake image or outperform one generation of artificial intelligence. It is to build systems that continue protecting people as the threats evolve. When individuals, businesses, and institutions have better ways to verify what they see and hear, we are doing more than improving technology. We are helping preserve the trust that makes digital communication possible.”