On-chain privacy: From "optional" to "essential"

Author: ChainUp Investment

1. Introduction

In 2025, on-chain privacy experienced a large-scale repricing event. Notably, driven by a resurgence of privacy awareness within the industry and significant advancements in cryptographic technology , Zcash witnessed substantial price discovery. These advancements included zero-knowledge proofs (ZKPs), multi-party computation (MPC), trusted execution environments (TEEs), and fully homomorphic encryption (FHE).

• ZKPs : A method for proving the validity of a statement without revealing any information other than the validity itself , enabling users to publicly share proofs of knowledge or ownership without disclosing details.

• MPC : A cryptographic protocol involving multiple parties collaboratively computing data by splitting it into " secret shards ." No single party can see the complete data.

• TEE : A hardware-based solution. It is a secure " black box " inside the processor used to isolate data as it is used.

• FHE : An encryption scheme that allows direct computation on encrypted data without decryption .

The market has shifted from “ anonymity ” to “ confidentiality ,” which is a functional necessity in transparent blockchains.

On-chain privacy attention surged in Q4 2025. Source: Dexu

1.1. The Privacy Paradox

The history of privacy cryptocurrencies dates back to 2012 when Bytecoin launched CryptoNote, providing ring signature technology, which was later used by Monero in 2014. In short, privacy is not a new concept in cryptocurrencies, but in its early stages, privacy cryptocurrencies were largely an ideological pursuit or circumvention tool, and a channel for bad actors to evade surveillance. The struggles with on-chain privacy in previous years can be attributed to three main factors: technological immaturity , fragmented liquidity , and regulatory hostility .

Historically, cryptographic techniques have faced scrutiny for their high latency and inefficiency . Today, the proliferation of developer tools (zkDSLs) like Cairo and backends like Halo2 has made ZKPs accessible to a wider range of developers. The trend of building zkVMs (zero-knowledge virtual machines) on standard instruction sets like RISC-V is making the technology scalable and composable across a variety of applications. MPC is no longer just for splitting private keys ; through MP-SPDZ , it supports arithmetic circuits (addition/multiplication) and Boolean circuits (XOR/AND), allowing for general-purpose computation. Advances in GPUs have further benefited these technologies; for example, the H100 and Blackwell B200 now support confidential computation, enabling AI models to run in TEEs. The biggest bottleneck in FHE, bootstrapping latency (i.e., the time to "refresh" noise in cryptographic computation to continue processing), has been reduced from approximately 50 milliseconds in 2021 to less than 1 millisecond in 2025, enabling real-time deployment of FHE cryptographic smart contracts.

zkVM iterations and performance, source: Succinct, Brevis

Furthermore, privacy is often isolated on specific blockchains, forcing users to leave their existing active ecosystems across chains to achieve anonymity, which is costly in terms of transaction fees and capital opportunity costs. Today, privacy protocols like Railgun can be directly integrated with DeFi applications, providing privacy as a shield against copy trading and MEV withdrawals. Boundless, Succinct, Brevis, and similar protocols provide ZKPs-as-a-Service for applications, while Arcium and Nillion help build privacy-preserving applications using MPC , and Phala and iExec compute application data within TEEs without leaving their blockchains. Finally, Zama and Octra enable applications to natively handle FHE computations .

Railgun TVL, Source: DefiLlama

In its early days, blockchain needed transparency to gain legitimacy . Genuine builders had to distance themselves from hackers, money launderers, and other bad actors. In this environment, privacy features were quickly seen as tools for dishonest actors. Projects like Tornado Cash, while attracting privacy-conscious users, left these users with funds mixed with illicit actors, unable to prove their innocence. The result was regulatory crackdown. Exchanges, in pursuit of operating licenses, froze funds from mixers and delisted questionable privacy tokens. Venture capitalists and institutional funds refused to hold them for fear of compliance officials. On-chain privacy became a “criminal” feature in the industry. Today, the economic sanctions against Tornado Cash have been lifted. The industry has coalesced around the concept of “ compliant privacy ,” designing “ visible data ” that allows users to decrypt their own fund origin transactions by providing auditors or regulators with a “ viewing key .” This approach can be seen in Tornado Cash and Zcash.

Significant impact of sanctions on Tornado Cash fund flows, Source: Dune

2. Current use cases for privacy technologies

Early setbacks don't mean privacy isn't important. Ask yourself a simple question: " Would you want your coffee-buying habit today to expose your entire 10-year investment history? " Most people would say no, but that's exactly what current blockchain setups do. As crypto legislation progresses and more institutions join the fray, these new institutional players are re-examining this issue. Fortunately, by 2025, the adoption of privacy technologies will be driven more by functional utility than ideology.

2.1. Block transactions

Leveraging a “ visible data ” design, Zcash’s shielded supply has increased from 12% in early 2025 to approximately 29% currently. This increased demand is due to a combination of factors, including increased speculative interest in the ZEC token and the natural desire to shield transactions from the public. The mechanism for shielding transactions is called the Commitment-Nullifier Scheme , where senders can submit shielded vaults to a pool. The network uses ZKPs to verify the submitted funds to prevent double-spending and creates a new shielded fund vault for the recipient.

ZEC supply is blocked on Zcash. Source: ZecHub

One of the fastest-growing sectors, crypto-neobanks, is actively exploring implementing privacy-focused transactions for their users, such as Fuse , Avici , and Privily . This is despite the fact that some protocols use different methods to mask on-chain transactions.

2.2. High-performance execution environment

Based on Total Value Locked (TVL), ZK-2 layer networks grew by 20% in 2025, offering a significantly cheaper execution environment compared to Ethereum's layer 1 networks. Layer 2 networks package all transactions on their network into a tiny data blob and send it to a sorter to generate proofs, which are then sent to the underlying layer 1 network for verification.

Applications of major ZK-2 layer networks: TVL variation trends, Source: DefiLlama

Today's ZK offers comprehensive built-in privacy features, such as privacy smart contracts on Aztec and ZKsync Interop , which unifies liquidity between the ZK chain and Ethereum.

2.3. MEV Protection

One of the most common use cases for privacy "hiding" is preventing Maximum Extractable Value (MEV). The transparent nature of blockchains allows predatory bots to view transactions in a public mempool before confirmation and engage in preemptive or "sandwich" transactions to extract profits. Flashbot SUAVE is decentralizing the block building process through an encrypted mempool, where transactions remain encrypted until the block builder commits to include them. Unichain has also introduced TEE-based block building to ensure that transactions on Layer 2 networks cannot be preempted.

Percentage of transactions landing at Flashbot Protect, Source: Dune

2.4. Other Use Cases

Aside from the main exceptions mentioned, developers are actively exploring the implementation of on-chain privacy on their applications for optimization and a better user experience.

• Order Book : Hyperliquid whales like James Wynn and Machi Big Brother frequently face liquidation hunts. While Hyperliquid's founders believe transparency provides a level playing field for market makers and leads to tighter spreads, the risk of being front-runners or trading against them is a significant net negative for whale traders. This creates an opportunity for Aster to offer privacy features such as hidden order books and a new Shield Mode launching by 2026.

• Identity : Certain activities, such as new bank account applications and initial coin offerings (ICOs), require verification of the applicant's identity. Protocols like idOS allow users to upload KYC information once and reuse it seamlessly across other compliant protocols; zkPass helps provide users with Web2 credentials without revealing details; World ID proves a user's identity through iris hashing; and ZKPassport verifies a user's identity without the information leaving the user's own device.

  • SEC Chairman Paul Atkins stated that many types of ICOs should not be considered securities and therefore fall outside the SEC's jurisdiction. His stance could trigger more ICO fundraising in the near future, thereby increasing the demand for crypto KYC (Know Your Customer) procedures.

• Cross-chain bridges : Throughout blockchain history, cross-chain bridges have been vulnerable to exploitation. For example, Ronin Bridge and Multichain suffered losses of $624 million and $126 million respectively due to private key leaks. ZK-based cross-chain bridges minimize trust assumptions , offer instant determinism once proofs are generated and verified, and are scalable and cost-efficient as transaction volume increases. Polyhedra Network uses zkBridge to connect over 30 chains and can be plugged into the LayerZero V2 stack as a "DVN".

• AI : ZK helps verify that the output is generated based on expected inputs and processed by a specific model. Giza enables unmanaged agents to execute complex DeFi strategies based on verified generated outputs. Phala uses Intel SGX enclaves to securely store sensitive information such as private keys in the AI ​​agent.

3. Core DeCC Ecosystem Classification

On-chain privacy typically refers to decentralized confidential computing networks (DeCC). While the market often categorizes protocols based on their underlying privacy technologies, each privacy stack has its trade-offs, and an increasing number of protocols are adopting hybrid approaches to their privacy solutions. Therefore, it is best to classify them as privacy blockchains, privacy middleware, and privacy applications.

Core DeCC Ecological Classification

3.1. Privacy Blockchain

The "privacy blockchain" category includes both Layer 1 and Layer 2 networks, where privacy mechanisms are embedded in the consensus or execution environment. A core challenge for these networks is the " cross-chain barrier ." Attracting users and liquidity to migrate from established blockchains is extremely difficult without a killer application to make the migration economically viable. Privacy Layer 1 network tokens are typically assigned a " Layer 1 network premium " because they are used as security collateral to protect the network and as gas tokens.

3.1.1 The Legacy and Evolution of Layer One Cyber ​​Privacy

Zcash has historically been positioned as the privacy-focused Bitcoin of its kind. The network features a dual-address system, allowing users to switch between public and private transactions, and includes a "view key" to decrypt transaction details for compliance purposes.

The protocol is transitioning from Proof-of-Work (PoW) consensus to a hybrid model called Crosslink , which will integrate Proof-of-Stake (PoS) elements in 2026, providing faster determinism than the initial probabilistic determinism of the Nakamoto consensus. Following the halving in November 2024, the next halving event is expected to occur in November 2028.

On the other hand, Monero maintains its default privacy approach, using ring signatures, stealth addresses, and ring CT to enforce each transaction. This design choice led most exchanges to delist XMR tokens from their platforms in 2024. Furthermore, Monero experienced several hashrate attacks from Qubic in 2025, resulting in a reorganization lasting 18 blocks and the erasure of approximately 118 confirmed transactions.

Secret Network is a privacy-preserving layer network based on a TEE (Threaded Execution Environment), built on the Cosmos SDK since 2020, and includes view keys for access control. Secret not only positions itself as a standalone chain but also provides TEE-as-a-service for EVM and IBC chains. The team is also focused on providing confidential computation in AI and exploring the integration of threshold-based confidential computing (FHE) into the network.

Canton Network is backed by Wall Street giants such as Goldman Sachs, JPMorgan Chase, Citi Ventures, Blackstone, BNY, Nasdaq, and S&P Global. It is a layer-one blockchain designed to introduce trillions of dollars in RWA (Real-World Assets) through a unique privacy feature called the Daml Ledger Model . Parties in the Daml ledger can only view a subset of the ledger connected to their subnet; this model allows verification only by the parties involved in the transaction, and unrelated parties are not actually aware of the transaction's existence .

Aleo is a ZK layer-1 network that uses the proprietary Rust-based language Leo to compile code into ZK circuits. Users generate proofs of transaction execution off-chain (or pay miners to generate them), and then only send the cryptographic proofs to the network.

Inco positions itself as a Layer 1 network for FHE , while also providing FHE-as-a-service to other chains through cross-chain bridges and messaging protocols. This same functionality enables the chain to serve deep liquidity without requiring the chain to build its own DeFi from scratch.

Octra is a high-performance FHE layer-1 network . Octra built its own proprietary cryptography from scratch, called Hypergraph FHE (HFHE) , which allows parallel processing during computation and achieved a peak throughput of 17,000 TPS on its testnet.

Mind Network utilizes restaking protocols such as EigenLayer to secure the FHE validator network. This protocol aims to create an end-to-end encrypted internet (HTTPZ) and enable AI agents to process encrypted data.

3.1.2. The ZK-Layer Two Network

ZKsync has expanded from simple scaling to implementing a range of comprehensive solutions, such as Prividium, ZKsync Interop, and Airbender . Prividium allows companies to execute transactions privately while still using Ethereum for final secure settlement. Airbender is a high-performance RISC-V zkVM prover that can generate ZK proofs in sub-seconds. ZKsync Interop allows users to provide collateral on the ZK chain and borrow assets on Ethereum.

Starknet leverages STARKs (Scalable Transparent Knowledge Arguments) for high throughput scaling and features native account abstraction . Each account on Starknet is a smart contract, allowing for the execution of invisible transactions using account contracts. The team also proposed Ztarknet , a layer-2 network that settles on Zcash, introducing a smart contract platform that benefits from Zcash's anonymity.

Aztec operates as a native privacy layer-2 network on Ethereum, using a UTXO-like ticketing system to process encrypted data and an account-based system to process public data. Aztec's Noir -based architecture relies on client-side proofs or privacy execution environments (PXEs), where users generate ZK proofs locally on their devices and then send them to the network.

Midnight operates as a partner chain of Cardano, utilizing Cardano's Stake Pool Operators (SPOs) for security while running its own execution layer. It is a ZK Layer 1 network based on TypeScript and selective disclosure capabilities . It uses ADA for secure staking, unmasked NIGHT tokens for governance and staking to generate Gas (DUST), and default-masked DUST as the Gas token.

Phala relies on TEEs such as Intel SGX to protect privacy. The protocol has shifted to an AI coprocessor mode , allowing AI agents to run within the TEE and manage private keys, and collaborates with Succinct and Conduit to migrate from the Polkadot parachain to the Ethereum Layer 2 network using the OP Succinct stack.

Fhenix is ​​the first fhEVM layer-2 network on Ethereum, bringing cryptographic computation to the Ethereum ecosystem. Transactions executed on this chain are protected by MEV because transaction inputs are encrypted in a mempool.

3.2. Privacy "Middleware"

Privacy middleware protocols operate on a Privacy-as-a-Service ( PaaS ) model, providing computational power for proof generation, encryption, or verification. Competition in this area is fierce regarding latency, cost efficiency, and network support.

Boundless , an "universal ZK computation layer" incubated by RISC Zero , is a decentralized ZK proof marketplace . It allows any blockchain or application to outsource heavy proof computations to Boundless.

Succinct Labs is a direct competitor to Boundless, positioning itself as a high-performance prover network. It adds dedicated circuitry to its zkVM (SP1) for common tasks such as hashing and signature verification, making proof generation faster and cheaper .

Brevis, as a ZK coprocessor , allows smart contracts to query historical data from any blockchain without trust. Now, Brevis extends to the general-purpose zkVM via Pico , where, in addition to pre-compiling for heavy workloads, the coprocessor can also be integrated as a dedicated circuit.

Arcium, as a performance-adjustable MPC solution, serves applications on any chain, although it uses Solana for staking, staking, and node coordination.

Nillion also provides high-performance MPC services for applications. Its Nil Message Compute (NMC) and Nil Confidential Compute (nilCC) enable fragmented data to be computed without sending messages to each other during the computation phase, while remaining secure within the TEE .

iExec RLC has been a long-standing dePIN protocol since 2017, providing cloud computing resources. Now, it shifts its focus to TEE-based confidential computing , allowing AI models to be trained or queried without revealing data inputs, and providing privacy tasks for chains like Ethereum and Arbitrum.

Marlin also underwent a major transformation, from a blockchain CDN to a confidential computing layer (Oyster) and the ZKP marketplace (Kalypso) built on top of its computing layer.

Zama is the leading FHE protocol for building fhEVM, TFHE-rs, and Concrete , which are used by protocols such as Fhenix and Inco. Zama also offers FHE as a service on existing public blockchains. With its recent acquisition of Kakarot, it also plans to integrate FHE into zkVM.

Cysic builds physical hardware ( ASICs ) to accelerate ZKP generation, reducing proof generation time from minutes to milliseconds. Users can request proof generation from ZK Air (consumer-grade) or ZK Pro (industrial-grade ASIC).

3.3. Privacy Applications

This is the largest category of privacy blockchains and privacy middleware, and the list in this article only shows a small fraction of them. The protocols here leverage ZK, MPC, TEE, or MPC to improve the user experience of their products. Successful applications abstract away the complexities of privacy protection and provide truly product-market fit solutions.

Tornado Cash was the original decentralized and tamper-proof mixer . The protocol was sanctioned by the U.S. Treasury Department in 2022, with the sanctions subsequently lifted in early 2025. Nevertheless, it remains a high-risk tool for compliant entities.

Railgun is widely recognized as having received endorsement from Vitalik Buterin. It offers a voluntary disclosure shielding trading solution that surpasses Tornado Cash by integrating users' " safes " with DeFi protocols such as Uniswap or Aave. Although its shielded assets are only around 20% of Tornado Cash's, it is still widely considered a potential competitor to Tornado Cash.

World (formerly Worldcoin) uses iris scanning to establish " proof of identity ," in which biometric data is encrypted, and only ZKP is sent to the network. World ID has become an effective tool for distinguishing between robots and AI.

zkPass uses a third-party TLS handshake to allow users to generate proof of their personal identity and media profile data , thereby enabling access to gated applications without revealing private information.

Privy enables users to seamlessly log in to decentralized applications using their email or Web2 accounts , creates MPC wallets for users, and splits the keys between the user's device and a security server. This essentially eliminates the cumbersome process of backing up mnemonic phrases and significantly improves the user experience.

Aster has partnered with Brevis to build its Aster Chain, offering privacy-focused transactions on top of its current hidden order book. The protocol roadmap indicates that Aster Chain is expected to launch in the first quarter of 2026.

Malda is a unified liquidity lending protocol that utilizes boundless proofs to manage users' lending positions across multiple blockchains .

Hibachi provides a high-frequency decentralized perpetual exchange and utilizes Succinct to prove its off-chain Central Limit Order Book (CLOB) for on-chain verification.

Giza introduces machine learning into smart contracts, allowing them to run validation outputs from anticipated AI models. This enables AI-driven DeFi strategies to execute on-chain without manipulation.

Sentient is a dedicated AI-layer network (powered by the Polygon CDK) designed to create an open AGI platform and reward contributors accordingly. AI model owners upload their proprietary AI models to the network and receive rewards based on usage. Models on the platform have cryptographic fingerprints to ensure that certain outputs are generated by a specific model. It also builds the Sentient Enclaves Framework, leveraging AWS Nitro Enclaves to enable confidential computations within AI models , shielding node operators from user prompts and the model's internal state.

4. Current Trends and Future Outlook

4.1.1. The Rise of Privacy Middleware

We are witnessing a shift from monolithic privacy chains to modular privacy layers. Protocols don't need to migrate to a privacy blockchain; instead, they can be deployed on any established blockchain, such as Ethereum and Solana, while accessing privacy services via smart contracts , thus minimizing the barrier to entry for accessing the protocol . Furthermore, with increasing demand for privacy features and growing privacy awareness in the industry, privacy middleware is the ultimate beneficiary, as running its own computationally intensive, confidential computing framework is economically unfeasible for many startup protocols.

Number of proofs requested and completed on Succinct, Source: Dune

4.1.2. Hybrid Solutions

Current privacy enhancement technologies have their own limitations . For example, ZKP cannot perform computations on encrypted data , MPC may be limited by latency when there are many participants, TEE may be compromised by fault injection and side-channel attacks (where attackers gain access to physical hardware), and in FHE computation, complex calculations may take longer and the risk of data corruption due to accumulated noise is higher. Therefore, more and more protocols tend to use a hybrid approach, combining their privacy tools or designing dedicated hardware to optimize computation.

4.1.3. Confidential and Verifiable AI

Morgan Stanley estimates that global AI-related capital expenditures will reach $3 trillion. With AI demand projected to expand in 2026, confidential and verifiable AI has become a major trend in 2025 and is expected to scale up in 2026. Confidential model training on sensitive data such as medical and financial records could be another significant milestone in the field of decentralized AI.

5. Summary

The era of privacy tokens lacking "view keys" may be coming to an end. The industry is betting that this " selective disclosure " approach will be accepted as a full compromise. If regulators later reject this approach, it could force networks to opt for "regulated permissioned chains" to achieve anonymity.

The maturation of privacy-enhancing technologies is key to unlocking trillions of dollars in traditional finance. Bonds, securities, and corporate payrolls cannot exist on a transparent blockchain. As these protocols prove robust by 2025, we anticipate the first major privacy-enhancing RWA pilots to launch on one of the aforementioned networks in 2026.

Google Trends on “Blockchain Privacy” over the past 5 years. Source: Google

While the hype surrounding blockchain privacy may cool down temporarily , demand for privacy features at the application layer is expected to grow steadily , significantly improving user experience and attracting a large non-crypto-native audience. This marks a turning point where on-chain privacy is shifting from " optional " to " essential ."