Aryan Nath
I am a first-year Master’s student in the Computer Science at Aarhus University, specialising Cybersecurity. Before that, I completed a BSc in Computer Science with a Minor in Mathematics at Ashoka University.
I am broadly interested in applied cryptography, anonymous credentials, and applied machine learning.
Research
-
Stable Private Intersection with Consistency
Abstract
Private set intersection (PSI) enables two mutually distrusting parties to learn extit{only} the intersection of their private sets. Malicious security for PSI, however, constrains extit{how} a party computes and not extit{which} inputs it computes on. A party running multiple PSI sessions can use different inputs each time or change its inputs across sessions to bias the output or infer more information from the intersection. Recent work binds parties to committed sets [Goel et al., USENIX Security 2026; Sun et al., CCS 2024, 2025] but gives the sender only extit{weak} input consistency. That is, the sender cannot inject elements but it can omit any elements from the intersection. Omitting elements causes (1) false negatives (i.e., an element in both sets is silently dropped from the intersection) and (2) biased outputs (i.e., the sender picks which intersection elements each receiver gets).
We introduce Stable Private Intersection with Input Consistency (SPICy), a framework that binds both parties to their exact committed input sets across an unbounded number of intersections. We define two notions of input consistency: (1) extit{strong} input consistency (SIC) wherein each party uses its full committed set, though the receiver’s output may still differ from the intersection of the committed sets; (2) extit{complete} input consistency (CIC) wherein the receiver’s output is the exact intersection of the committed sets. SPICy builds on MiniPSI [Rosulek and Trieu, CCS 2021], the fastest known small-set PSI protocol, through a new primitive, the Mutual Oblivious PRF (MOPRF). In an MOPRF, both parties evaluate a PRF under their own keys, and the receiver learns the sender’s output only on elements held by both parties. SPICy achieves CIC for both the sender and receiver and requires a single round of online communication. We implement SPICy in Rust and evaluate it end-to-end on deployed PSI applications. For sets of sizes between , SPICy has – times online runtime overhead and times communication overhead compared to MiniPSI.
-
An Explainable Deep-Clustering Approach to Assess Climate Vulnerability in India
Teaching
-
Undergraduate teaching assistant
Information Security.
-
Undergraduate teaching assistant
Computer Networks.
-
Lab assistant
Computer Networks.
-
Undergraduate teaching assistant
Data Structures and Algorithms.
-
Undergraduate teaching assistant
Discrete Mathematics.
-
Teaching Assistant for Mathematics
Number Theory.
Awards
-
Gold Medal in Computer Science
Awarded to the top-ranked student of the BSc (Hons) Computer Science class.
-
J.N. Tata Endowment Scholarship
For graduate study at Aarhus University.
-
Third Place, ICH Hackathon
As the only student-led team in the final round, for our work on explainable deep clustering for climate vulnerability in India.
Misc
- Swimming: I swam competitively at the national level during my school years in India.
- Music: I play the guitar and sing, and performed in a band during my bachelor’s.
- Languages: English (fluent), Hindi (native), Danish (beginner; currently taking lessons at A2B Sprogskole).