Zara on building verifiable privacy

Zara Javar holds a PhD in Computer Science from the University of Victoria, where her thesis, "Formal Algebraic Reasoning about Compression Functions Security" focused on the formal verification of hash function security. She brings 8+ years of experience spanning theoretical cryptography research and production-oriented protocol design, specializing in SNARKs/STARKs, fully homomorphic encryption, and multi-party computation.

Before joining the Interfold, she was a Security Apprentice on Nethermind's Audit & Research team, working on ZK proof system efficiency and smart contract security audits. She also serves as adjunct faculty at the New York Institute of Technology, teaching advanced cryptography.

• Name: Zara Javar
• Role at the Interfold: Senior Cryptography Engineer
• Area of expertise: zk-SNARKs and zk-STARKs, threshold fully homomorphic encryption (FHE), secure multi-party computation (MPC), lattice-based cryptography (RLWE), elliptic curve cryptography and polynomial commitments, formal security verification, automated security verification frameworks
• Links: @zarajavar / linkedin

Hi Zara! What’s your background, and what drew you to cryptography / confidential computation?

My path started in pure math. I've always loved problem-solving and the elegance of mathematical reasoning. But what really drew me into cryptography was the chance to apply that math to real-world privacy challenges. During my master's, I worked on privacy-preserving data mining, investigating how differential privacy, FHE, and MPC could be integrated to protect sensitive datasets like hospital records. I assessed their strengths and weaknesses in the context of distributed data, which gave me a solid foundation in understanding how these primitives complement each other. That's when I realized that the problems I was solving mathematically had a direct, tangible impact on people's privacy. 

My PhD pulled me deeper into the theory side, where I built frameworks for formal algebraic reasoning about hash function security - essentially, automated security verification. Instead of just analyzing a single hash function, I designed systems that could automatically reason about compression functions and verify their security properties in both the ideal cipher and random oracle models. That experience taught me how to think about security guarantees systematically, which is exactly the mindset needed for building protocols like PV-TBFV. 

Confidential computation felt like the natural place to bring all of that together: it's where the formal guarantees actually have to hold up against real adversaries and real systems.

What do you enjoy the most about working on the Interfold?

What excites me most is that we're building infrastructure where parties can coordinate on shared computation without revealing their individual inputs - and we're using the most promising cryptographic tools available to get the strongest possible security guarantees. We combine zero-knowledge proofs for public verifiability, fully homomorphic encryption for computation on encrypted data, and multi-party computation for distributed key management - all working together.

Blockchain is the ideal verification and accountability layer for this because it provides a trustless, public ledger where anyone can verify the protocol's outputs. The integrity and privacy guarantees come from cryptography and economics rather than from trusting a piece of hardware - and the blockchain gives us the transparency and accountability layer to make that work. If a ciphernode misbehaves or aborts, the on-chain evidence is there to trigger slashing, creating real economic consequences for malicious behavior.

I think we're just scratching the surface of what's possible at the intersection of privacy and blockchain. As public verifiability becomes the norm, we'll see a new generation of applications - private voting, sealed-bid auctions, confidential DeFi - that combine the transparency of blockchain with the privacy guarantees of advanced cryptography. Getting to build that future, from research design through to production, is rare and incredibly exciting.

What technical problem or area are you currently focused on?

I've been working on the design of PV-TBFV with the Interfold team. It's a collaborative effort that's become the cryptographic core of the Interfold's Encrypted Execution Environment (E3). The protocol combines FHE, MPC, and zero-knowledge proofs into one pipeline: distributed key generation, encryption, and threshold decryption, each step verifiable on-chain. Concretely, that's meant building out eight ZK circuits in Noir and a recursive proof aggregation scheme, plus the security analysis underneath it – RLWE parameterization, smudging noise bounds, and ensuring the commitments stay consistent across every circuit.

We've already applied this to private voting use cases, and we're expanding to others like sealed-bid auctions - any scenario where parties need to compute on encrypted inputs without revealing them. Right now, we're also working on optimizing the circuits by adding a folding scheme, which will significantly reduce proving costs and make the protocol even more practical for on-chain deployment.

How do you see confidential computing evolving over the next 5 years?

I expect we'll see public verifiability become the default expectation for confidential computation, rather than a nice-to-have. Right now, proving that encrypted computation was done correctly is still expensive, but FHE schemes are getting more efficient and ZK proving systems are improving rapidly. We're seeing constant-size proofs, faster proving times, and better developer tooling. As those trends continue, the overhead of public verifiability will drop to the point where it's practical for a much wider range of applications.

Another major shift is the move toward post-quantum secure systems. Quantum computers pose a real threat to classical cryptography, and the community is already responding. We're seeing lattice-based ZK proofs emerge as a promising direction-systems like Symphony (a scalable SNARK from lattice-based folding) and LaBRADOR (a transparent lattice-based proof system with sublinear proof sizes) are concrete examples of where the field is heading. Lattice-based cryptographic assumptions are widely regarded as the most promising candidates for post-quantum cryptography, and integrating them into ZK systems is a natural next step.

I also think we'll see FHE, MPC, and ZK proofs combined more deliberately, the way we're doing with threshold FHE, rather than treated as separate, competing tools. The synergy between these primitives - using ZK to verify FHE operations, using MPC to distribute trust, and now building all of this on post-quantum foundations - is where the most powerful and future-proof applications will emerge.

Can you share some advice for developers or researchers interested in privacy-preserving technologies?

Get your hands on an implementation early - write a small ZK circuit, or implement a basic threshold scheme - because the gap between the theory and what actually has to be true for a real system is where most of the interesting problems live. And don't be afraid of the formal security side; being able to state precisely what a protocol does and doesn't guarantee is what lets you trust it in production.

Technical Spotlight

Title: PV-TBFV: Publicly Verifiable Threshold BFV for Confidential Coordination

Description: PV-TBFV is a publicly verifiable threshold BFV protocol - threshold fully homomorphic encryption combined with MPC and zero-knowledge proofs so that every phase of the process can be checked by anyone, not just trusted by the parties running it. This covers distributed key generation, encryption, and threshold decryption, with publicly verifiable secret sharing underpinning the key generation step and an automated BFV parameter search tool (built in Rust) to choose secure, efficient parameters for a given deployment.

A key innovation is the BFV-on-BFV PVSS layer: instead of introducing a new encryption scheme for share distribution, we encrypt Shamir shares under a second BFV instance. This exploits the RLWE polynomial structure - each polynomial coefficient of a threshold-BFV secret share can be encoded as a plaintext coefficient in the second BFV scheme - so an encrypted share sent to one party encrypts an entire polynomial of share coefficients at once.

The protocol is implemented as eight ZK circuits in Noir covering distributed key generation, public-key aggregation, encryption, and threshold decryption, with recursive proof aggregation so the whole pipeline can be verified on-chain efficiently rather than checking each step independently. Alongside the circuits, the work included a full correctness and security analysis across every phase -RLWE parameterization, smudging noise bounds, cross-circuit commitment consistency, and the Fiat-Shamir transformations used to make the proofs non-interactive. The protocol is proven secure in the Universal Composability (UC) framework, with formal definitions of robustness, public verifiability, and identifiable abort.

This is the cryptographic core of what the Interfold calls an Encrypted Execution Environment (E3): independent parties can submit encrypted inputs, have a computation run over them, and get a verifiably correct result, without any single party, or any piece of trusted hardware, ever seeing the underlying data. That's the mechanism that lets the integrity and privacy guarantees come from cryptography and economics rather than from attestations, which is the core bet the Interfold is built on.

Quick Picks

Favourite research paper: Urban and Rambaud's "Robust Multiparty Computation from Threshold Encryption Based on RLWE"

Favourite tool / repo: Noir – the ability to write ZK circuits in a Rust-like language and target UltraHonk on BN254 makes it surprisingly practical to build complex protocols like PV-TBFV.

One concept you think more people should understand better: The difference between public verifiability and privacy. A common misconception is that you have to sacrifice one for the other. PV-TBFV shows that's not true - you can have both. Zero-knowledge proofs let you verify correctness without revealing inputs, and threshold FHE lets you compute on encrypted data. The combination is powerful, and I think more people should understand that these aren't trade-offs – they're complementary tools.