By Jack Imel and Kimberly Mann Bruch —
Neha Sangwan, a postdoctoral scholar at UC San Diego’s Halıcıoğlu Data Science Institute (HDSI), has been selected as the recipient of the prestigious 2025 IEEE Information Theory Society Thomas M. Cover Dissertation Award for tackling how to keep our connected devices secure. Her research examined a major problem: what happens when some devices (such as computers and cameras) in a network are broken or have been hacked by an adversary.
Sangwan’s award-winning dissertation, Communication with Byzantine Users, explored a prominent security challenge to modern digital systems, the presence of malfunctioning or intentionally malicious components. She worked on the project while earning her doctorate degree at the Tata Institute of Fundamental Research in Mumbai. She is now focused on an array of projects with UC San Diego professors Tara Javidi and Arya Mazumdar. Javidi is a professor in the Department of Electrical and Computer Engineering at the UC San Diego Jacobs School of Engineering and co-director of the Center for Machine-Intelligence, Computing and Security at UC San Diego. Mazumdar is a professor at HDSI, which is part of the School of Computing, Information and Data Sciences (SCIDS) at UC San Diego and associate/deputy director of NSF AI Institute TILOS and site-lead of NSF Data Science Foundations Institute EnCORE
“We are pleased that Sangwan was presented with this extremely prestigious and competitive award, which is presented each year for an excellent doctoral dissertation, across the globe, that contributes significantly to the foundations of mathematics in the field of information sciences,” said Javidi, who is the Jerzy (George) Lewak Chair and a professor in the Jacobs School of Engineering at UC San Diego and is also affiliated with HDSI. “Both Arya and I look forward to continuing our work with her on statistical learning theory, and information acquisition for quantum detection as well as modern AI systems,” said Javidi.
“Sangwan is a thorough researcher,” added Mazumdar. “She is able to break complicated research questions into small parts and systematically tackle them; she has flourished in HDSI.”
As for the study she conducted related to the award, Sangwan explained that systems of multiple internet-connected devices are ubiquitous in our present era. Smart city infrastructure, security systems, smart homes, wearable technology, cars with features like GPS navigation or self-driving and blockchain networks with uses ranging from election security to healthcare record management are all examples of such interconnected digital systems – distributed networks with the connected devices called nodes.
Ideally, communication between these nodes is completely reliable and secure, but this is not always the case. Malfunctioning or hacked nodes can send erroneous messages across the network, leading to dysfunction in systems responsible for protecting private medical data or keeping homes secure.
Sangwan said that rogue nodes like this are known as “byzantine nodes,” named for “The Byzantine Generals Problem,” which is a classic riddle in distributed computing that poses the question: how can properly functioning nodes know what information to trust when other nodes in the network may be malicious or unreliable?
The consequences of byzantine nodes extend far beyond the digital realm and have the potential to imperil human life and limb. For example, malfunctioning cameras or radar on a self-driving car could result in the anti-collision system activating erroneously, or not activating at all.
“This is not a hypothetical scenario,” Sangwan said. “Since October 2024, the National Highway Traffic Safety Administration has been investigating malfunctioning self-driving systems in connection with crashes, one of which resulted in a fatality.”
Sangwan’s dissertation and continued research aim to better understand how to create more efficient, fault-tolerant distributed networks. Her goal is for networks to be able to cope with the presence of malfunctioning or malicious byzantine nodes to ensure the safety, security, privacy and general well-being of end-users.
To comprehend the brilliance of Sangwan’s work, one must first understand the foundational building blocks of fault tolerance in distributed networks: Byzantine Broadcast and Byzantine Multiple-Access. She explained that a single, potentially byzantine message-sending node communicates with multiple honest receiving nodes in Byzantine Broadcast. Any of these nodes could be malicious and may not follow the communication protocol.
“The use of the name ‘Byzantine’ comes from a hypothetical – though realistic – scenario set in the Byzantine Empire. Imagine a Byzantine army besieging an Ottoman city, with generals sending messages to their lieutenants commanding the frontline,” Sangwan illustrated. “For the sake of simplicity, consider one general communicating with two honest lieutenants. This general will either command the lieutenants to retreat or advance. But, the general may secretly be a traitor and intentionally send different messages to each lieutenant, disrupting the attack.”
In order to maintain coordination on the battlefield, the two lieutenants must agree on the same course of action, whether the commander is honest or not. The same is true for maintaining coordination in a distributed network: the receiving nodes must all agree on the same message, even when the sending node may be an unreliable, “Byzantine” node.
On the other hand, Byzantine Multiple Broadcast involves multiple sender nodes, each transmitting a separate message to one receiver node. This is known as a multiple access channel (MAC). When all senders are honest, the receiver nodes must reliably decode the messages of all senders. But, if some of the senders are malicious Byzantine (called a Byzantine-MAC), the receiver must continue to decode the messages of honest nodes correctly, or in the case of less critical systems, flag the presence of such Byzantine nodes. In this scenario, the former requirement is known as “reliable communication,” while the latter is known as “authenticated communication.”
“A Byzantine-MAC may be used to model the up-link communication in a smart home where wireless devices communicate with a central server and some of them may be malfunctioning or hacked,” explained Sangwan. “Imagine a home where a security system, climate control, lightbulbs, speakers and a desktop computer are all part of a distributed network and one of these devices – such as the computer – has been hacked. In such a situation, we would like to ensure graceful degradation of services depending on the criticality of the application.”
She explained that in this example, more critical devices like the security system are required to lock the system so it can continue functioning with reliable communication. On the other hand, the least critical devices like the climate control or music system can shut down on detecting the presence of malicious devices, satisfying authenticated communication.
“This approach allows networks to prioritize what’s most important,” Sangwan said. “Critical safety systems stay online and secure, while less important systems can safely disconnect when threats are detected.”
Sangwan completed her Ph.D. at the Tata Institute of Fundamental Research in Mumbai under the supervision of Vinod Prabhakaran in 2023. She received an MS.c. degree from Chennai Mathematical Institute in 2017 and a B.E. degree from the Delhi College of Engineering in 2013.
