COLLECTIVE SECURITY
PhD, Georgia Institute of Technology
Distinguished Paper Award, ACM CCS 2023
Best Demo Award, IEEE ICDCS 2024
IEEE Distinguished Lecturer, IEEE VTS
the partnership
Agent systems become collective systems the moment their actions cross boundaries. Balaji Palanisamy’s work on blockchains, privacy, security, and scalable cloud and edge systems gives that collective layer a durable footing.
The partnership focuses on making group behavior resistant to compromised participants, poisoned shared resources, and coordination failures — so no single agent can silently destabilize the fabric around it.
the research
Three collective-security directions translate Dr. Palanisamy’s work on privacy, distributed systems, and blockchain resilience into the agent fabric.
DIRECTION 01
Emergent Coordination & Swarm Security
Study how agent collectives coordinate safely when local decisions compound into system-wide behavior, incentives, and attack surfaces.
Collective Security · coordination resilience · swarm integrity
DIRECTION 02
Compromised- & Byzantine-Agent Resilience
Build runtime mechanisms that keep the collective trustworthy even when some agents are faulty, adversarial, or strategically deceptive.
Byzantine resilience · distributed trust · fault containment
DIRECTION 03
Agent Memory & Shared-Resource Poisoning
Protect shared memory, state, and coordination resources from subtle contamination that can spread through an entire agent fleet.
Memory safety · shared-state integrity · privacy-preserving systems
supporting research
Privacy-Preserving System Design
Design distributed systems that preserve privacy while still supporting useful coordination, accountability, and scalable analysis.
Privacy · data security · collective trust
Scalable Cloud & Edge Resource Management
Manage shared compute and data resources so collective agent workloads remain available, efficient, and resistant to contention.
Cloud scale · edge systems · shared-resource resilience
why it matters
“The riskiest failures in agentic AI won’t come from one agent going rogue. They’ll come from many well-behaved agents adding up to something nobody authorized. That’s a systems theory problem before it’s a security product problem — and it’s exactly the class of problem Balaji has spent his career getting right.”
— Sandeep Gopisetty, Founder & CEO, Algedonic