Dr. Arun Balaji Buduru, Founding Head, Usable Security Group, gave a talk on Leveraging AGI Models for Proactively Protecting Users and Devices on 24 June.
Here is the summary of the talk:
Cyber systems, including the Internet of Things (IoT), are widely being used to vastly improve operational efficiencies and reduce costs in vital sectors, including finance, transportation, defence and healthcare. The majority of today’s economy has become more digitised over the last three decades due to significant advancements in hardware costs and computer efficiency. It is crucial to recall that the widespread usage of devices and services has led to the creation of vast volumes of rich user data, which can be used as a weapon to defraud consumers and jeopardise device security.
Majorly, current cyber defence systems have two main problems:
(1) They fail to take into account the development of LLM-based models as a significant threat vector
(2) They fail to continuously adapt to the dynamic environment required to counter sophisticated cyberattacks.
As a result, the efficacy of the security solutions created has significantly decreased. The speaker discussed some of his work that addressed the aforementioned issues in this lecture, specifically
(1) source attribution of generative models in audio deepfake detection by behavioural analysis and
(2) Keystroke identification with user and keyboard-neutral features.
Dr. Arun Balaji Buduru is currently heading the Usable Security Group at IIIT-Delhi and is an Associate Professor at Dept. of CSE and HCD in IIIT-Delhi. He received his Ph.D in Computer Science, specializing in Information Assurance, at Arizona State University in 2016. His research interests include Usable Security and Privacy, Affective AI and Computational Linguistics. He received his B.E. degree in Computer Science in 2011 from Anna University at Chennai, India. He worked as a research intern as part of Cisco IoT Architecture Group, San Jose. His Usable Security Group at IIIT-Delhi currently focuses on developing user-centric solutions for problems in the domains of audio deepfake detection, speech emotion detection, speaker verification, and malware behavioral analysis.
His research works are published in several prestigious journals and conferences such as Springer Scientific Reports, ACM AsiaCCS, CHI, IJCAI, CHI, ACL, AAAI ICWSM, NAACL, INTERSPEECH, ICASSP, AAMAS, ACM HT, IEEE ICME, ACM WebSci. He believes in developing efficient algorithms with learning-theoretic guarantees, whose effectiveness can be validated in practical settings. As a consequence, he actively engages with many governmental agencies and private companies such as DRDO, ISRO, ANRF, IDS, NTRO, Delhi Police, CISCO, ABB, DCM Tech by taking up research projects for real-world deployment.
Website: http://faculty.iiitd.ac.in/~arunb/
June 2026