Dedeepya Yarra and the rise of Trustworthy AI: Where Machine Learning meets cybersecurity
Dedeepya Yarra's interdisciplinary research interest is the intersection of machine-learning research and cybersecurity, MLOps and cloud technologies.
Artificial Intelligence is going beyond the ability to make accurate predictions or automate complex tasks to demonstrate success. With the advent of more and more machine-learning applications in cybersecurity, healthcare, business processes, and other critical areas, the concerns of security, resilience, reliability and responsible deployment are also gaining traction. Intelligent systems are thus protected from new and emerging threats and can continue to operate in a reliable way through the convergence of machine learning, cybersecurity, cloud infrastructure and operational practices in the emerging field of trustworthy AI.
Dedeepya Yarra's interdisciplinary research interest is the intersection of machine-learning research and cybersecurity, MLOps and cloud technologies. She also is dedicated to the study of the behavior of intelligent systems subject to attacks, when deployed at scale and integrated into real-world infrastructure, a shift in focus from the performance of the models to how the systems behave.
She has worked on adversarial machine learning, focusing on adversarial black-box and white-box approaches. They explore how models can be threatened, manipulated or tested in various manners, and gain understanding of vulnerabilities that may prove more challenging to uncover. Moreover, she has investigated the use of machine-learning based intrusion detection with the Kitsune model, using AI methods to bridge the gap between the real-world problem of discovering potentially advanced threats in the cybersecurity landscape.
Meanwhile, her studies have branched out from cybersecurity. Yarra has explored deep-learning techniques for segmenting customers and creating personalised marketing strategies, looking at how behavioural data can be used to inform more personalised marketing strategies. She has also researched optimisation methods with respect to the prediction of breast cancer, which is a healthcare-related problem that can be solved with computational approaches. In addition, her academic research has spanned a range of current topics in machine learning, such as image sets, and Generative Adversarial Neural Networks.
These experiences will be part of the reasons why she feels her work is linked to the conversation about trustworthy AI. Adversarial machine learning moves security to the heart of development and evaluation of intelligent systems, rather than an afterthought tacked onto the end. As a result, knowledge of how models react to a variety of attack scenarios may be an essential component of evaluating the reliability of the models.
However, the trust in AI goes beyond individual models and involves the context within which AI systems function. That's where Yarra's fascination with MLOps is most applicable. The MLOps field combines machine learning, software engineering and infrastructure approaches to solve a problem that occurs after making an AI model. All of these steps are needed to sustain a system over time: deployment, monitoring, automation, security, updates, and scalability.
Yarra has collaborated with technologies and practices such as Kubernetes, Docker, Cloud Platforms, Infrastructure-as-Code, Automation and Observability in this area. These spaces serve as indicators of a significant operational aspect of AI. A model can be effective during development, but in production, it needs to be reliable depending on the ability to deploy, monitor and maintain the surrounding infrastructure consistently. This has led to the emergence of MLOps in AI development, highlighting the need for trustworthy systems, which goes beyond complex algorithms.
She is also interested in the applications of technological concepts in various fields as reflected in her wider investigation into new technologies. Yarra created a framework on the use of the blockchain in children's education and presented research for the same at IEEE CCWC 2024. Although this is not the same research undertaken for her cybersecurity and machine-learning projects, this study follows a persistent trend in her research: studying how emerging technologies can be leveraged to solve a real problem.
In the meantime, her activities in cybersecurity research have spread to professional and academic activities. Yarra was invited to the University of North Texas Cyber Security Month to speak about black-box and white-box attacks and the impact of adversarial machine learning to cyber security. She has also given research presentations on deep learning and customer segmentation, which has helped in the debate on the growing use of AI.
All of these are aspects of work that highlight the critical need for trustworthy AI in more impactful applications as AI becomes more pervasive. Even capable models can be compromised by security issues, and poor infrastructure can pose reliability and operational issues. In turn, building AI systems that can be secured, monitored, evaluated, and maintained is increasingly intertwined with the larger issue of whether such systems can be trusted.
Yarra's work falls at the intersection of these changing fields of machine learning, cybersecurity and infrastructure. She works on adversarial attacks in her research, and systems and environments that enable reliable AI operation in her MLOps interests. Meanwhile, her wider research interests reflect the growing prominence of AI and its related technologies in various sectors.
The focus is now moving from the mere creation of intelligent systems to the creation of systems that can handle real-world conditions, as AI technology advances. During that shift, security, resilience and operational reliability are increasingly a part of the discussion. In its interdisciplinary efforts, Yarra illustrates this, identifying one of the main issues that the next generation of AI will have to address: not only to be intelligent, but to be reliable enough to be trusted.
For breaking news and live news updates, like us on Facebook or follow us on Twitter and Instagram. Read more on Latest India News on India.com.