In short
OAST, or out-of-band application security testing, finds vulnerabilities that give no visible response. A test payload makes the target server contact an outside server the tester controls, usually by DNS or HTTP. If the callback arrives, the bug is real. Burp Collaborator and Interactsh are the best-known OAST tools, and attackers use them too.
Key takeaways
- What it is
- A testing method that confirms blind bugs, such as blind SSRF or blind SQL injection, by watching for callbacks to an external server.
- How it works
- Each probe carries a unique subdomain of the callback server. If the target resolves or fetches it, the OAST server logs the interaction and the tester polls for it.
- Why it matters
- The same callbacks let attackers find exploitable servers at scale and even move data out over DNS. The callback has to leave your network to work.
- What Odin Research saw
- 1,794 callback endpoints in Monograph 17, 757 of them confirmed or suspected nodes. Most are ordinary testing servers; 18 hosts share an address with other offensive tools.
- What to do first
- Stop internal servers from resolving or calling the internet directly. Route their DNS through Protective DNS and alert on callback-style lookups.
What is OAST?
OAST, or out-of-band application security testing, is a way to find vulnerabilities that show no visible sign in the application's response. The test payload tells the target server to contact an outside server, usually by DNS or HTTP. If that callback arrives, the hidden flaw is confirmed.
PortSwigger, which makes Burp Suite, puts it this way: OAST "uses external servers to see otherwise invisible vulnerabilities". PortSwigger says it was a pioneer of the method with Burp Collaborator.
Some of the most serious server bugs are silent. A blind server-side request forgery (SSRF), a blind SQL injection or an XML external entity (XXE) flaw may run in the back end without changing the page. Error messages do not appear, and timing may not change. A callback is often the only proof.
OAST is a normal part of penetration testing and bug bounty work. It is also dual-use: attackers rely on exactly the same signal to find servers they can exploit.
How OAST works
An OAST setup has a callback server and a tester's client. PortSwigger's documentation for Burp Collaborator describes the loop:
- The scanner puts a payload into a request. The payload contains a unique subdomain of the callback server's domain.
- If the target is vulnerable, it looks up or contacts that subdomain.
- The callback server records the DNS query, HTTP request or other interaction.
- The scanner polls the callback server and matches each interaction to the payload that caused it.
The callback server runs its own authoritative DNS name server, so every lookup under its domain reaches it, even if the target cannot open a direct connection. ProjectDiscovery's Interactsh, a popular open-source OAST server, captures DNS, HTTP(S), SMTP(S) and LDAP interactions. Self-hosted Interactsh servers can also capture FTP and SMB, and the SMB option records NetNTLMv2 hashes.
OAST compared with DAST and SAST. PortSwigger describes dynamic testing (DAST) as sending payloads and reading responses, which misses blind and asynchronous bugs. Static testing (SAST) reads code without running it. OAST adds an out-of-band channel to dynamic testing, and PortSwigger says it rarely produces false positives.
OAST tools and public callback servers
Three tools dominate the OAST servers Odin Research sees on the internet:
- Burp Collaborator, PortSwigger's service used by Burp Suite Professional and Burp Suite DAST. PortSwigger also documents how to deploy a private Collaborator server.
- Interactsh, ProjectDiscovery's MIT-licensed server and client, written in Go. Its client ships with a set of public default servers under several
oast.*domains, and the project recommends self-hosting if those change or go down. - DNSLog, a lighter DNS-focused callback collector that Odin mostly sees on Chinese cloud providers.
Self-hosting an Interactsh server takes a domain with its own name-server records and a small always-on server, according to its README. That low cost is why so many private callback servers exist, for testers and attackers alike.
How attackers use OAST
Attackers use OAST for the same reason testers do: to find out, at scale, which servers are really vulnerable. A mass scan sends tagged probes to thousands of hosts. The callback server then lists exactly which ones answered.
Common uses include:
- Confirming blind flaws before spending effort on a full exploit, for example JNDI lookups of the kind used against Log4Shell.
- Moving data out over DNS. Small chunks of data can ride inside DNS labels sent to a callback domain, a form of data exfiltration that many firewalls let through.
- Feeding automation. Monograph 17 models how an AI offense engine, a blind probe dispatcher, a callback collector and an implant listener could form one pipeline. Odin presents this as a model, not an observed operation.
The weak point for the attacker is that the callback must leave your network. That makes egress control and DNS logging the strongest defenses.
What Odin Research observed
- 1,794 endpoints, mostly testers. Monograph 17 mapped 1,794 out-of-band callback endpoints on 1,052 hosts. Of these, 757 are confirmed or suspected nodes: 638 Interactsh, 109 private Burp Collaborator and 8 DNSLog, plus 2 offensive workbenches.
- Company matters. On 18 hosts, a callback listener shared an address with another offensive framework, such as Sliver, VShell, Cobalt Strike, ARTEX, the ARL recon panel or the ezXSS manager. Eight of those are corroborated.
- Attacker files. Open web directories held 40 callback records. Eighteen sat on three confirmed offensive workbenches, inside Python DNS exfiltration scripts and Java exploit templates. These show intent and tooling, not a working attack on any target.
- A cloud cluster. On 18 November 2026, Odin first observed a DNSLog collector, a VShell listener, a CyberStrikeAI engine and a HexStrikeAI bridge on four separate servers in one Alibaba Cloud range. Nothing links them, so they stay on watchlists.
- Look-alikes. Monograph 13 found 356 AWS load-balancer endpoints labeled as Interactsh servers. Of these, 189 shared the exact TLS fingerprint of AWS-hosted Interactsh servers, so only a certificate issuer and application-layer check told them apart.
- Leaked tokens. Monograph 09 counted 143 per-session OAST and Telegram tokens among secrets found on exposed attacker hosts.
The lesson: a callback server on its own is a dual-use testing tool. Raise priority when it shares a host with C2 or exploit tooling.
How to detect OAST callbacks
- Server DNS lookups. A database or middleware server has no reason to resolve a random label under a public callback zone. Alert on lookups of the Interactsh
oast.*zones, Burp Collaborator domains and DNSLog domains from server subnets. - Random-looking labels. Monograph 17 recommends alerting on 26 to 36 character random labels in server DNS traffic, which is how correlation IDs look.
- Telltale headers. Interactsh servers name themselves in response headers, and Burp Collaborator servers do too. These help you spot a callback server in your logs.
- Odd outbound protocols. Treat anonymous LDAP binds to the internet, and database functions such as
xp_dirtreeorUTL_HTTPreaching out, as high priority.
Monograph 17 publishes ten Suricata signatures for callback DNS lookups, OAST server headers and polling, plus a Sentinel KQL hunt, three Sigma rules and four YARA rules. Monograph 13 adds Suricata signatures for the Interactsh banner and client polling. Use them as hunting rules, since your own testers trigger them too. All are in the Defense Kit.
How to defend against OAST-based attacks
- Cut server egress. Block direct outbound DNS, HTTP, HTTPS, SMTP and LDAP from web, middleware and database subnets, except to explicit allowlists.
- Use Protective DNS. Make servers resolve only through internal forwarders that log queries and can sinkhole callback zones.
- Turn off risky database features such as
xp_dirtreeandUTL_HTTPwhere you do not need them. - Register your own tests. Allowlist approved scanner addresses and private callback domains before each test window, so real alerts stand out. Track the rest as false-positive candidates, not threats.
- Fix the root cause. Every confirmed callback is a real bug. Route it into vulnerability management.
OAST tools compared
| Tool | Who makes it | Protocols | What Odin saw |
|---|---|---|---|
| Interactsh | ProjectDiscovery (MIT, Go) | DNS, HTTP(S), SMTP(S), LDAP; FTP and SMB self-hosted | 638 confirmed or suspected nodes, mostly Amazon and DigitalOcean |
| Burp Collaborator | PortSwigger | Payload subdomains polled by Burp | 109 private servers, mostly Amazon and DigitalOcean |
| DNSLog | Several community projects | DNS-focused | 8 nodes, mostly Tencent and Alibaba in China and Hong Kong |
| ezXSS | Blind XSS manager | Blind XSS callbacks | Seen next to callback listeners in Monograph 17 |
Frequently asked questions about OAST
What is OAST?
OAST stands for out-of-band application security testing. It finds vulnerabilities that give no visible response by making the target contact an external server the tester controls, usually over DNS or HTTP.
What does OAST stand for?
Out-of-band application security testing. Out-of-band means the evidence arrives through a separate channel, such as a DNS lookup, rather than in the application's own response.
Is OAST malicious?
No. OAST is a standard testing method used by penetration testers, bug bounty hunters and scanners. Attackers use the same technique, so callbacks from your servers should be explained, not ignored.
What is the difference between OAST and DAST?
DAST sends payloads and reads the responses. OAST adds an out-of-band channel, so it can confirm blind bugs that DAST alone would miss.
What is an OAST domain?
It is the domain of a callback server, such as the public Interactsh oast.* domains or a Burp Collaborator domain. Test payloads use unique subdomains of it so each callback can be matched to its probe.
Interactsh vs Burp Collaborator: which should I use?
Both work. Burp Collaborator is built into Burp Suite Professional and Burp Suite DAST. Interactsh is free, MIT-licensed and easy to self-host, and it adds LDAP, FTP and SMB capture on self-hosted servers.
How do I detect OAST callbacks in my network?
Log DNS from server subnets and alert on lookups of public callback zones or long random labels. Watch for callback server headers and outbound LDAP. Odin Research publishes Suricata, Sigma, YARA and KQL rules for this.
Should I block every OAST server?
No. Most are run by legitimate testers. Block direct egress from your servers instead, and raise priority only when a callback server shares a host with C2 or exploit tooling.
Sources
- PortSwigger, "Out-of-band application security testing (OAST)". portswigger.net/burp/application-security-testing/oast (checked 21 November 2026)
- PortSwigger, Burp Collaborator documentation. portswigger.net/burp/documentation/collaborator (checked 21 November 2026)
- ProjectDiscovery, Interactsh repository, GitHub. github.com/projectdiscovery/interactsh (checked 21 November 2026)
Related tools and terms
What is JNDI injection?
JNDI injection makes a Java application look up a remote object, usually over LDAP or RMI, and load attacker code. Log4Shell is the best-known example.
TechniqueWhat is HVNC?
Hidden VNC (HVNC) creates a second, invisible desktop on the victim's computer so an attacker can use the victim's browser sessions unseen.
VulnerabilityWhat is Ghostcat?
Ghostcat (CVE-2020-1938) is a flaw in the Apache Tomcat AJP connector that lets an attacker read files and, in some setups, run code.
TechniqueWhat is JARM?
JARM is an open TLS server fingerprinting method from Salesforce. It hashes how a server answers a set of crafted TLS handshakes.
TechniqueWhat is xp_cmdshell?
xp_cmdshell is a SQL Server extended stored procedure that runs operating system commands. Attackers enable it to move from database access to the host.
TechniqueWhat is DCSync?
DCSync abuses the Active Directory replication protocol to ask a domain controller for password hashes, as if the caller were another domain controller.
Odin Research
Get the rules, not just the definition
Every Odin Research report ships with verified indicators and detection rules you can load today.