What is data detection and response?
Data detection and response (DDR) is a cloud security discipline that continuously monitors sensitive data stores in real time to detect unauthorized access, behavioral anomalies, and active exfiltration attempts. By analyzing cloud access logs, database transactions, and workload runtime events, DDR triggers automated containment workflows to stop data breaches before exfiltration completes.
A few behaviors that typically trigger a detection look like this.
A service account querying sensitive customer tables outside its microservice scope.
A developer role bulk-downloading objects from cloud storage at 2:00 AM from an unfamiliar IP address.
Rapid creation, unencryption, or export of database snapshots to external cloud accounts.
None of these requires a misconfiguration. Each one uses access that already exists, which is why posture checks on their own tend to miss them.
Incident Response Plan Template
Prepare your security team to detect, contain, and remediate cloud threats and data breach incidents with our customizable response framework.

How DDR differs from DSPM, CDR, and DLP
Data loss prevention (DLP) blocks specific actions based on static rules, like "do not send credit card numbers to a public domain." It works at network egress and endpoint boundaries, where regex patterns and file fingerprints can be checked.
DDR watches a different place. It tracks data access inside cloud data stores and responds to abnormal behavior, even when every request is a valid API call.
Two other categories sit close by, and data detection and response depends on both. Data security posture management (DSPM) inventories and classifies data at rest, which is the posture prerequisite DDR builds on. Cloud detection and response (CDR) covers control-plane and compute threats, such as a hijacked workload or cloud account.
The table below compares all four. The eBPF telemetry it mentions is explained in the next section.
| Capability | DDR | DSPM | CDR | DLP |
|---|---|---|---|---|
| Primary Scope | Dynamic data access, queries, and exfiltration | Static data inventory, classification, and posture | Cloud control plane, identity, and compute runtime | Endpoint files, email egress, and web gateways |
| Operational State | Data in motion and data in use | Data at rest and dormant snapshots | Workload processes and control plane APIs | Network egress and endpoint transfer boundaries |
| Telemetry Ingested | Object storage logs, database audit logs, eBPF | Cloud metadata, disk snapshots, storage scanning | CloudTrail, VPC flow logs, audit logs, runtime sensor | OS kernel hooks, browser extensions, proxy inspection |
| Detection Method | Behavioral anomaly detection, query baselines | Policy misconfigurations, excessive permissions | Threat intelligence, MITRE ATT&CK rules, anomalies | Static regex matching, file fingerprinting, keyword dictionaries |
| Primary Outcome | Rapid containment of active data exfiltration | Remediation of exposed databases and toxic access paths | Neutralization of compromised workloads and cloud accounts | Blocking unauthorized file transfers at network boundaries |
The categories work best together. DSPM tells you which stores matter, CDR watches the infrastructure around them, and DDR watches the data itself. If you already run endpoint detection and response (EDR), keep it, because DDR adds the data context that makes endpoint findings easier to act on.
How data detection and response works
DDR builds on the sensitive-data inventory that posture tools already maintain. It then layers two telemetry sources and a correlation step on top, so each detection reflects both what happened and what it touched.
Control-plane log ingestion
The control plane is the first telemetry layer. DSPM already knows which stores hold sensitive data, so DDR ingests their access logs from AWS CloudTrail data events, Azure Storage Analytics, Google Cloud Audit Logs, and Snowflake or BigQuery query streams.
From those logs, DDR learns what normal looks like for each data store. A bulk download from an account that typically reads single records breaks that baseline. So does access from a location the user has never connected from.
This layer is a practical place to start data detection and response, because the logs already exist in every major cloud. The real work is turning on data-level events. CloudTrail trails and event data stores log management events but not data events by default, so you need to enable them and route them somewhere they can be baselined.
Workload and container runtime telemetry
Control-plane logs have a blind spot. They record API calls against a data service, but they don't see what happens inside a compromised host. An attacker with stolen host credentials can read a mounted volume, query a local database, or stage files without triggering a single control-plane configuration event.
Runtime sensors close that gap. eBPF is a Linux kernel technology that lets small, sandboxed programs observe system activity, including network calls, process events, and file operations, safely and without modifying kernel code or loading a kernel module. Kernel-level eBPF sensors use it to inspect in-memory queries and local file reads on virtual machines, containers, and Kubernetes nodes.
In practice, the two layers confirm each other. A spike in local file reads on a workload, followed by new outbound connections, tells a clearer story than either signal alone.
Contextual correlation and prioritization
Raw anomaly detection creates noise. DDR cuts it by joining identity entitlements from cloud infrastructure entitlement management (CIEM) with the sensitivity label on the data. That join separates a real threat from a nightly extract, transform, load (ETL) job reading large volumes on schedule.
Sensitivity then sets the response. An unauthorized user accessing a file with personally identifiable information (PII) gets escalated immediately. The same access pattern on a public marketing document gets logged but not alerted.
Automated containment
DDR can contain a threat without waiting for a human to review the alert. Say someone starts downloading sensitive files from an IP address in a country where your organization has no employees. Within seconds, DDR can block the transfer, revoke the user's access token, and alert the security team.
The scope stays narrow on purpose. Revoking one session token through cloud provider APIs stops the download without breaking the pipelines that share the same data store.
Automating these first steps of cloud incident response gives analysts their time back for investigation.
Watch 5-min demo
Data detection and response teams can explore Wiz Defend’s approach to cloud threat containment in a five-minute recorded demo.

Data detection and response use cases
The value of data detection and response shows up in specific attack patterns. Each scenario below is a data exfiltration detection problem where the attacker's access looks legitimate at the API level.
Double extortion ransomware
Ransomware operators now exfiltrate data before encrypting it, so they can threaten a leak even after you restore from backups. A CISA advisory on Interlock ransomware describes actors who encrypt systems only after exfiltrating data.
That staging window is where DDR helps. Bulk reads across many tables, compression into large archives, and unusual egress volumes all show up before encryption starts.
When DDR revokes the staging identity's access at that point, the attacker loses the copy that makes extortion work. Recovery then becomes a restore problem instead of a negotiation.
Compromised identities and non-human accounts
A stolen service account key or session token produces valid, authorized API calls over legitimate HTTPS. Insider threats look the same way, because the identity is real and the permissions are already granted.
Only behavioral baselines reveal the anomaly. Volume, tables touched, time of day, and source location show when an authorized credential starts acting like someone else. Non-human accounts are often the easiest to baseline, since a service account's job rarely changes.
Generative AI and RAG pipeline data leakage
Retrieval-augmented generation (RAG) pipelines pull production data into places it never used to go.
A typical leak starts with production customer records copied into vector databases like Pinecone or Milvus. Another path sends records to an unauthorized third-party large language model (LLM) API.
DDR flags the unusual read volume from the pipeline's identity and the new egress destination. Your team can then cut that path before sensitive records end up in an index or an outside model.
Connecting data detection to cloud context
A data detection is only actionable when it links back to its cloud cause. If an alert can't show the role, policy, trust relationship, or network path that allowed access, teams end up stitching evidence together by hand. Responders also need to know whether the host behind it is internet-facing, carries an exploitable vulnerability, or runs with overprivileged permissions.
That context is rarely optional. Wiz Research's Cloud Data Security Snapshot found that 54% of cloud environments have exposed VMs and serverless instances containing sensitive information.
Wiz approaches cloud data detection and response from that angle. Wiz DSPM flags anomalous access to sensitive data stores, and the Wiz Security Graph connects that access to toxic combinations of exposure, vulnerabilities, and permissions. Wiz Defend then contains the threat, for example by revoking an identity and access management (IAM) role or isolating a VM.
Wiz provides visibility across cloud, SaaS, on-prem, and hybrid infrastructure through the Sensor Workload Scanner. Explore how Wiz DSPM and Wiz Defend connect real-time data detection to cloud context, or get a demo to see it in your own environment.
Data Detection and Response with Wiz
Data detection and response combines sensitive data insights with cloud context, and a Wiz demo lets teams explore how Wiz supports investigation and containment.