Notice
This article was written based on one paper publicly available as of August 19, 2026. DANA NOTES analysis is also included.
Reference Paper
Zhu, Z., Lam, C., Frolov, A., Miers, I., & Liu, Z. (2026). Zero-knowledge cloud analytics. In ACM SIGCOMM 2026 Conference. Association for Computing Machinery.
What This Article Covers
Companies accumulate large amounts of data that are difficult to obtain externally. Information such as customer service usage patterns, network quality, security events, operational efficiency, and regional demand can support a company’s internal decision-making while also holding high value for other companies.
Two conditions must be met for this data to be used across companies.
Information providers need to protect personal information and trade secrets while making use of the value of their data. Purchasing companies need to obtain the results they need quickly and also secure evidence that those results were calculated accurately from actual data.
Zero-Knowledge Cloud Analytics, presented at ACM SIGCOMM 2026, is a study that addresses this issue through a cryptographic approach.
The researchers propose zk-Analytics, which processes raw logs within the analytics provider’s system and provides analytics results and cryptographic proofs to external verifiers. Instead of directly examining raw logs and intermediate analytics states, external verifiers can verify whether the results were calculated accurately from previously committed data.
If this structure develops for real-world enterprise environments, the way companies use data could also change.
Alongside transactions in which companies transfer entire raw datasets, transactions that provide “verifiable insights” calculated from raw data could grow into a new type of data product.
Key Terms to Know
Zero-Knowledge Proof (ZKP)
A cryptographic technique that proves a predefined computation was performed correctly while limiting the amount of private information used in the computation that is revealed externally.
Cloud Analytics
The analysis of large-scale data collected in cloud environments, such as server logs, network traffic, service usage, and security events.
Verifier
An external party that checks whether analytics results were calculated correctly. In zk-Analytics, this can include customers, regulators, auditors, and insurers.
Cloud Analytics Results Need a Process for Verifying Trust
Cloud service companies collect various types of data during their operations.
Video streaming services measure bitrate, resolution, video startup time, and rebuffering. Internet service providers measure actual network speed and throughput. Security service companies analyze security events and detection results generated across customer networks and systems.
These metrics are used for service-level agreements (SLAs), industry benchmarks, regulatory compliance, and security audits.
In many current cloud analytics environments, the analytics provider that collects and analyzes the data also provides the final results. Customers, regulators, and auditors need a way to confirm that those results were calculated accurately from actual data.
Raw logs can contain sensitive information such as IP addresses, cookies, request URLs, network configurations, and routing policies. They can also contain information about a company’s internal operations and customer behavior.
Zhu et al. (2026) describe the issue that arises here as a trust gap.
The purpose of zk-Analytics is to bridge this trust gap with cryptographic proofs.
zk-Analytics Connects Logs to Query Results in a Verifiable Structure
zk-Analytics consists of three main stages.
1. Records Are Linked When Data Is Generated
When logs are generated by servers, routers, and various devices, data sources organize them into batches.
Each batch of logs is linked through a hash chain. Because newly generated records are connected to previous records, the order and continuity of the logs can later be verified.
At regular intervals, the data source publishes a cryptographic digest of these records to a public transparency log.
This record becomes a reference point for confirming which logs subsequent analytics originated from.
2. The Analytics Provider Aggregates the Data
The logs are delivered to the analytics provider and organized into fixed-size units called epochs.
The analytics provider aggregates the raw logs into forms such as sums, averages, distributions, and frequencies. Data structures such as histograms, hash tables, and Count-Min Sketches are also used.
Zero-Knowledge Proofs are generated during this process as well.
The proofs are designed to verify that the data in each epoch is linked to the previously committed logs and that it was processed according to the predefined aggregation logic.
The researchers also applied a distributed processing architecture that divides computationally intensive aggregation work across multiple servers.
3. Results and Proofs Are Provided for the Required Questions
When an external verifier requests a specific analysis, the query engine retrieves the relevant epochs.
For example, queries could include:
- What was the average network throughput during a specific period?
- What level of playback quality did a specific streaming service achieve?
- How many times was a specific security threat detected during a certain period?
The query engine generates a Zero-Knowledge Proof together with the calculation result.
The verifier receives the result and proof and checks whether the computation was performed correctly. In the paper’s experiments, verification of already generated aggregation proofs and query proofs generally took tens of milliseconds.
Who Sees the Raw Data, and Who Verifies the Results?
Understanding zk-Analytics requires distinguishing the roles of three participants.
Data Source
The entity that generates raw logs, such as a server, router, or user device.
The paper assumes that data sources generate and record accurate logs according to predefined measurement methods.
Analytics Provider
The entity that receives the raw logs, stores them, aggregates them, and processes queries.
The analytics provider can see the raw logs. At the same time, the accuracy of its aggregation and query computations is subject to verification through cryptographic proofs.
Verifier
An external party that verifies the reliability of analytics results, such as a customer, auditor, regulator, or insurer.
The verifier examines analytics results and cryptographic proofs instead of raw logs and intermediate analytics states.
The privacy scope of zk-Analytics therefore focuses on protecting raw logs and intermediate analytics states from external verifiers while verifying the accuracy of computation results.
The disclosure scope of the analytics results themselves also needs to be managed within this structure. The researchers propose managing allowable queries through mechanisms such as approved query lists or runtime access control.
Initial Applications Include Streaming, Telecommunications, and Security
The researchers present three main use cases.
Comparing Video Streaming Service Quality
Streaming services collect Quality of Experience (QoE) metrics such as bitrate, resolution, video startup latency, and rebuffering during playback.
Analyzing data from multiple providers makes it possible to compare service quality and each provider’s position within the industry.
An analytics provider can manage the detailed QoE data of individual streaming services while externally providing results such as the performance level achieved by a particular provider.
zk-Analytics adds a proof that allows the accuracy of the computation to be verified.
Verifying Network Quality for Telecommunications Providers
Internet service providers measure regional connectivity, actual speed, and throughput.
This data can be used for service quality assessments and regulatory compliance.
Regulators or customers may need to confirm whether actual network performance in a particular region matches the level reported by the provider.
With zk-Analytics, internal network information and detailed traffic data can remain within the provider’s management scope while performance results and verification proofs are provided externally.
Verifying Security Service Detection Performance
Managed security service providers analyze their customers’ DNS, HTTP, and network connection records.
This data is used to create security metrics such as the number of connections to malicious destinations and the number of major detections.
These metrics can also be used in compliance audits and cyber-insurance underwriting.
Raw logs can contain hostnames, user information, and internal network structures. With zk-Analytics, auditors, regulators, and insurers can review the required security metrics and computational proofs.
Enterprise Data Use Requires Extensive Preparation
From a business perspective, this research raises another important question.
How can other companies use data held by a company?
Enterprise data use currently requires extensive preparation.
Information providers need to protect personal information, trade secrets, customer information, and internal operational information. They also manage the scope of data provision, purpose of use, retention period, access rights, and contractual terms.
For these reasons, many types of data are subject to restricted external use.
Purchasing companies face another issue.
When the desired data cannot be obtained directly, they need to create the necessary information themselves. They collect data from multiple sources, align the formats, store it, clean it, analyze it, and then review the results.
In activities such as market research, expansion into new regions, investment decisions, and competitive analysis, this preparation can require significant time and cost.
In many cases, what companies actually want is an answer needed for decision-making rather than an entire data file.
For example, a company considering expansion into a new region may need information such as:
- At what times is service demand concentrated?
- What level of network quality does a particular region have?
- What usage patterns appear within a specific customer group?
- What types of security risks occur frequently in a particular industry?
- Where are delays concentrated within the supply chain?
Verifiable analytics technology shows the possibility of turning these questions into transaction items of their own.
Information Providers Can Turn “Answers” into Products
From the perspective of an information provider, this structure could create a new way to use the economic value of data while maintaining control over the raw data.
A company can continue managing its raw data within its internal systems while providing externally requested analytics results.
For example, a single set of network data held by a telecommunications provider could produce several information products:
- Regional average network quality
- Traffic patterns by time of day
- Changes in the use of specific service types
- Connection stability metrics
- The percentage of regions meeting a specific quality standard
The same raw data can be used as multiple types of insight products.
When cryptographic verification is added, buyers can confirm whether those figures were generated according to the predefined data and calculation rules.
Information providers can generate ongoing revenue from their data while managing the scope of direct raw-data transfers.
The value of data can expand from an asset used in a single sale into an information asset that can repeatedly answer different questions.
Purchasing Companies Can Obtain the Information They Need More Quickly
For purchasing companies, reducing analytics preparation time can become an important source of value.
Using external data directly requires companies to understand the data structure, prepare a storage environment, clean the necessary fields, create queries, and review the results.
As verifiable insight transactions become established, purchasing companies can choose to purchase information based on the specific questions they need answered.
For example, if a company wants to know:
“What percentage of cases met the service quality standard in a specific region over the past six months?”
it can choose, alongside obtaining the complete raw logs and calculating the figure directly, to receive the result together with a verification proof.
This structure can reduce the time required to obtain information for market research and strategic planning.
When multiple companies use the same definitions and calculation standards, the speed of comparative analysis can also increase.
Buyers can expand the ways they use data by paying for the answers they need at the time they need them, as well as for ownership of the data itself.
The Enterprise Data Transaction Market Could Also Grow
If verifiable analytics technology becomes established in real-world industries, the size of the enterprise data transaction market could also grow.
Companies currently hold large amounts of data with high value whose external use remains limited.
Data involving personal information, customer behavior, operational methods, network structures, and security events can also be directly connected to a company’s competitiveness.
If companies can continue managing raw data internally while providing results to specific questions, information that has been difficult to use in the market could become a new type of product.
The transaction scope could extend across multiple forms of information:
- Aggregate statistics
- Industry benchmarks
- Market indicators
- Operational efficiency metrics
- Security risk indicators
- Service quality indicators
- Verification results for whether specific conditions are met
Information providers can increase the number of data products they can offer, while buyers can select the information they need more quickly.
As the range of tradable information expands, there could also be room for new data providers, analytics providers, and verification service providers to emerge.
The enterprise data market could expand from a raw-data sales market into a market that also trades verifiable information and insights.
This is DANA NOTES’ business interpretation, extending one step beyond the direct conclusions of the zk-Analytics paper.
The Criteria for Valuing Data Assets Could Also Change
The value of enterprise data is assessed based on factors such as scale, scarcity, freshness, and usability.
As the market for verifiable insights grows, additional criteria could be added.
What important business questions can this data answer?
To what extent can the computation behind those answers be provided in a verifiable form?
For example, even a company holding a large amount of data may have a limited range of monetization opportunities if it is difficult to create questions that external companies can use.
A company that can continuously generate information that matters to decision-making within a particular industry can achieve higher value from its data.
The competitiveness of data companies could also expand beyond data volume to include the value of the questions, the quality of the results, and verifiability.
The Largest Current Challenge Is the Cost of Proof Generation
For zk-Analytics to be used in real-world enterprise environments, performance and cost need to develop together.
The researchers evaluated the system using Google Cluster data, CAIDA network data, and vehicle emissions data.
In experiments involving 131,072 logs, ZK-based aggregation required approximately one million times (10⁶×) more processing time than conventional processing. ZK-based query processing required approximately 10,000 to 100,000 times (10⁴–10⁵×) more processing time.
Peak memory usage on aggregation servers was approximately 31–34 MB for conventional processing and approximately 9.4–9.5 GB for ZK processing.
The researchers separated online log commitment from proof generation to manage the computational cost.
Lightweight hash commitments are performed when the data is generated, while computationally intensive aggregation and proof generation are processed offline.
The system also uses distributed processing across multiple servers.
In experiments that increased the number of aggregation servers from one to eight, proof generation time was reduced by more than seven times.
When 131,072 logs were processed across eight aggregation servers, histogram- and hash-table-based aggregation proofs were generated within approximately 1.5 hours, while Count-Min Sketch processing completed within approximately four hours.
Verification of already generated proofs, however, generally took less than 100 ms.
Given the current performance characteristics, early applications can be explored in areas such as periodic performance reporting, regulatory submissions, security audits, and insurance underwriting, where the reliability of results carries high economic value.
DANA NOTES Commentary
The long-term business value of this research can be found in creating a new transaction model for internal enterprise information that has been difficult to trade, beyond the Zero-Knowledge technology itself.
Companies accumulate large amounts of data. That data includes information such as customer behavior, service usage patterns, operational efficiency, network conditions, and security risks that can also be valuable to other companies’ decision-making.
As verifiable analytics technology develops sufficiently, information providers can choose, alongside direct data transfer, a model in which they receive questions about their data and provide verifiable answers.
Buyers can use a model in which they purchase the insights they need alongside acquiring a complete dataset and analyzing it from the beginning.
If this structure becomes established across industries, some of the data currently retained within companies could be transformed into information products that can be traded in the market.
This is where the growth potential of the enterprise data transaction market emerges.
The market can grow through an increase in the types of information that can be traded, as well as through an increase in the amount of data entering the market.
The value of enterprise data could also be assessed through new criteria.
Alongside how much data a company holds:
What important questions can it answer?
How reliably can those answers be provided?
These two factors could become important elements in determining the competitiveness of enterprise data assets.
Zero-Knowledge Cloud Analytics remains a research-stage system. At the same time, this study shows the possibility that, as data protection and verification technologies advance, enterprise data could expand from an asset that is held into an insight asset that can be sold repeatedly.
What to Watch Next
The first variable is the pace at which Zero-Knowledge Proof generation costs decline.
The current prototype requires substantial computing resources for aggregation and query proof generation. Improvements in zkVM and ZKP performance could become important factors determining the pricing structure and scope of real-world services.
The second variable is deployment in real-world enterprise environments.
The paper evaluates the technology using real-world and synthetic datasets. As cloud providers, telecommunications companies, security companies, and data analytics providers begin applying it in actual operating environments, its economics and operating structure can be assessed more concretely.
The third variable is standardization of enterprise data products.
For the same question to carry the same meaning across different companies, data definitions, calculation methods, measurement standards, and verification rules need to be established together. As standardization progresses, it can become easier to create markets that compare results from multiple companies using the same criteria.
The fourth variable is the scope of use in audit and regulatory frameworks.
If cryptographic proofs are used as formal evidence in SLA verification, regulatory reporting, security audits, and insurance underwriting, demand for verifiable analytics could also expand.
The fifth variable is whether verifiable insights actually become tradable data products.
It will be important to observe whether aggregate statistics, industry benchmarks, risk indicators, service quality indicators, and verification results for specific conditions are formed into independent products.
If Zero-Knowledge Cloud Analytics becomes established in real-world industries, the enterprise data market could expand into a market that transfers data while also providing trusted information at the moment it is needed.
