The Security and Privacy Research Group (SyPy) at IIIT Hyderabad systematically uncovers architectural weaknesses in AI pipelines, blockchain networks, mobile ecosystems and distributed systems that underpin digital finance, identity and decision-making processes. Led by Prof. Ankit Gangwal, the group employs adversarial thinking to probe edge cases where code assumptions fail, preventing exploits that could cascade across interconnected platforms affecting millions. Operating within the Centre for Security, Theory and Algorithmic Research (CSTAR), SyPy fosters interdisciplinary approaches blending cryptography, systems analysis and algorithmic design to engineer resilience from foundational levels.
For CISOs building next-generation infrastructure, SyPy’s work offers a model for proactive vulnerability discovery in high-stakes domains where failures carry economic and societal costs, such as compromised payment ledgers or misdiagnosing AI models. In India’s expanding digital economy, the group’s emphasis on scalable defences aligns with national priorities for secure UPI scaling, sovereign AI stacks and privacy-preserving federated learning.
Model poisoning and watermarking secure AI intellectual property
SyPy researchers investigate model poisoning, where adversarial data subtly corrupts training pipelines, causing AI systems to fail selectively in production while passing validation checks, particularly dangerous in healthcare diagnostics or financial fraud detection. They also develop watermarking techniques embedding verifiable provenance signals into large language models, enabling creators to prove ownership against replication by competitors or thieves. Model merging—combining pre-trained foundations for efficiency—receives scrutiny for emergent vulnerabilities arising from unexamined interactions, guiding safer hybrid deployments.
Enterprises deploying generative AI must adopt these methodologies to safeguard investments in custom fine-tunes, ensuring model integrity through poisoning-resistant training and forensic traceability that withstands legal or commercial disputes.
Blockchain anomaly detection and malware behavioural analysis
In blockchain and cryptocurrency ecosystems, SyPy analyses transaction flows for fraud, laundering and manipulation patterns in pseudonymous ledgers lacking central remediation, designing architectures resilient to consensus attacks and oracle exploits. Malware research deconstructs adaptive strains that hide within legitimate processes, extracting behavioural signatures for predictive defence that anticipates evasion tactics over signature matching. Mobile security testing exposes data leakage via excessive permissions, inter-app pathways and background exfiltration, critical for platforms handling biometric unlocks and financial transactions.
These capabilities position SyPy as a vanguard for India’s blockchain ambitions, from CBDC pilots to DeFi platforms, while equipping mobile-first enterprises to harden against silent compromises that erode user trust.
Adversarial mindset drives research and talent pipeline
Prof. Gangwal cultivates a lab culture of ‘thinking like attackers’, training students to reverse-engineer assumptions and simulate exploits, producing researchers equipped for emerging threats in technologies not yet mature. Publications in top venues validate the group’s rigour, but its enduring impact lies in alumni shaping national cybersecurity postures across industry and academia.
For technology leaders, SyPy exemplifies academic-industry symbiosis where foundational research translates to deployable defences, recommending similar investments in adversarial red-teaming and university partnerships to build sovereign capability.
