AI Platform Flags Fraudulent Android Apps; Beats 3,000 Teams for Rs 3 Lakh

Fraudulent Android APKs have become a favored vector for banking fraud and malware delivery, and manual reverse engineering cannot keep pace with the volume of suspicious files. A student-built platform that coordinates autonomous AI agents across reverse engineering, static analysis, and dynamic testing offers a glimpse of how security teams could triage threats faster and more consistently.

Chandigarh University announced on September 11, 2026, through a PRNewswire release, that its four-member student team, Team Fi, secured second place in the national-level CyberShield Hackathon, co-organized by Bank of India and IIT Hyderabad. The team — Danish Verma, Rahul Jaluthria, Varun Gupta, and Avneet Kaur, all third-year BE-CSE (IBM-Cyber Security) students — won Rs 3 lakh for building an AI-powered cybersecurity platform designed to assist in the analysis and investigation of suspicious Android applications. According to the university, the platform harnesses generative AI for automated reverse engineering, static and dynamic analysis, and risk scoring of fraudulent mobile applications (APKs) and malware.

“

Chandigarh University announced on September 11, 2026, through a PRNewswire release, that its four-member student team, Team Fi, secured second place in the national-level CyberShield Hackathon, co-organized by Bank of India and IIT Hyderabad.

The competition was substantial. It attracted over 3,000 team registrations from universities and higher educational institutions across India. Of these, 72 teams involving 232 participants presented solutions and prototypes at IIT Hyderabad. Team Fi, representing the Centre for Privacy and Security in Emerging Technologies (CPSET) at Chandigarh University's Apex Institute of Technology-Computer Science Engineering (AIT-CSE), advanced among nine finalists for the first problem statement before being named runner-up.

The announcement lands amid India's rapidly expanding digital payments and mobile banking ecosystem. Android's open application distribution model makes it relatively easy for malicious or fraudulent APKs to reach users outside official app stores, including through phishing links and social-engineering campaigns. Banks and financial institutions increasingly need automated ways to triage suspicious applications at scale, because manual reverse engineering is slow, expensive, and limited by a shortage of skilled malware analysts. A successful AI-assisted pipeline could help investigators prioritize APKs by risk, reduce time-to-detection, and flag behavior patterns that static rules miss.

The team's stated approach is more ambitious than a conventional single-model classifier. The students describe an 'agentic AI infrastructure' that moves beyond a static analysis pipeline, exploring how autonomous AI agents can coordinate different stages of the investigation. In practice, this implies that separate agents might handle decompilation, string extraction, permission analysis, network-behavior simulation, code-flow review, and final risk aggregation. The idea is attractive because security analysis is multi-step and context-dependent; an orchestration layer that can reason about partial findings and request deeper analysis could outperform fixed pipelines. However, the release provides no benchmarks, false-positive rates, detection latency, or independent validation, so the prototype should be treated as an early-stage demonstration rather than a proven security product.

For the banking sector, the most immediate value is talent validation and fresh research. Bank of India's participation as co-organizer signals that financial institutions are actively scouting AI-driven fraud detection beyond traditional rule engines. If such tools mature, they could integrate with threat-intelligence workflows and possibly be embedded into bank app stores or customer-security teams. Yet production deployment would require rigorous testing against obfuscated and adversarial APKs, compliance with data-privacy rules, and guardrails around autonomous AI actions, especially if agents are allowed to execute sandboxed code or make quarantine recommendations.