Written by Alan Winters
As artificial intelligence models grow more sophisticated, so do the threats against them. Intellectual property theft, adversarial attacks, and model extraction are no longer hypothetical risks as they are active challenges for companies pushing the boundaries of AI. At the center of the fight to secure these systems is Param Popat, a machine learning Engineer at Apple. His AI security and work change how companies protect their most valuable digital assets.
“Security in AI is not just about protecting assets, it is about ensuring the integrity of systems that billions of people rely on,” he explains. His work, from a patent pending on safeguarding AI models to advanced simulation technologies, reflects a nuanced approach to tackling these critical challenges.
The Rise of AI and Its Vulnerabilities
The increasing ubiquity of AI has brought with it a paradox. The same advancements that make AI indispensable also expose it to vulnerabilities. Model stealing, for instance, allows adversaries to replicate companies’ intellectual property without authorization. The barriers to entry for attackers are lower than ever. With open-source tools and computational power widely available, even small-scale actors can attempt to exploit these systems.
For tech giants like Apple, the stakes are extraordinarily high. The theft or compromise of AI models could undermine years of research, erode user trust, and result in significant financial losses. With the global AI market projected to exceed $500 billion by 2025, the urgency for robust security measures cannot be overstated. Innovators like Param Popat play a vital role in addressing these vulnerabilities.
Param Popat’s Path to Apple
Born in India, Param Popat’s early fascination with computer science laid the foundation for an illustrious career. He earned a Master of Science in Computer Science from Columbia University, where he specialized in AI applications for edge devices, achieving a GPA of 3.98. His academic work demonstrated his ability to balance theoretical research with practical solutions.
His professional journey began with research internships at Bosch and AI Zwei, where he focused on model optimization and adversarial defense mechanisms. At Bosch, he developed a patented system to detect and protect against model-stealing attacks, creating a new revenue stream for AI security solutions. “The problem was not just technical,” he reflects. “Understanding the business implications of stolen models and creating a robust and scalable solution.”
Since joining Apple in 2021, Param Popat has expanded his focus, working on a spectrum of projects ranging from gesture recognition for the Apple Watch to photorealistic 3D simulations. These initiatives highlight his ability to translate advanced research into practical, impactful technologies.
A Game-Changing Patent in AI Security
Param Popat’s most notable contribution to date is his pending patent, “A Method to Prevent Capturing of Models in an Artificial Intelligence-Based System.” This introduces a system to safeguard AI models against reverse engineering, addressing a critical gap in AI security frameworks.
Unlike traditional encryption methods, his approach secures the model, making it significantly harder for attackers to exploit. “The goal was to create a mechanism that doesn’t just detect breaches but actively prevents them,” he explains. This patent could set a new standard for AI security, ensuring companies can confidently deploy models.
Applications Beyond the Patent
Param Popat’s patent extends beyond security. At Apple, he has developed a photorealistic 3D indoor simulator capable of training reinforcement learning agents in diverse virtual environments. This project, which employs advanced 3D Gaussian Splatting, enables realistic scenario testing for AI Agents and Systems, making it a valuable tool for AI development.
His contributions to Apple Watch’s AssistiveTouch gesture recognition system demonstrate his versatility. He delivered a solution that operates in real time while conserving battery life by blending technical sophistication with user-centric design through spatiotemporal models.
Balancing Innovation and Security
The dual imperatives of innovation and security often create tension, particularly in fields like AI. Param Popat’s approach prioritizes adaptability and ethical considerations. “AI systems need to evolve alongside the threats they face,” he asserts. This philosophy drives his work on privacy-preserving machine learning models, which ensure systems remain secure without compromising user trust.
Ethical challenges, such as balancing transparency with privacy, add another layer of complexity. Param Popat acknowledges these difficulties and emphasizes aligning technical solutions with societal expectations.
The Future of AI Security
The challenges for AI security are expected to intensify. The rise of generative AI and large language models introduces new risks that demand innovative defenses. Param Popat remains optimistic, emphasizing the importance of building trust in AI technologies and their governance.
The societal imperative in AI underpins everything from healthcare to financial markets. Securing these technologies is not just a technical challenge. Param Popat’s work embodies innovation and security, offering solutions that address some of the most pressing threats in AI today. As AI becomes increasingly integrated into daily life, experts like Param Popat are shaping a future where technology remains impactful and secure.
