Detections
Explore public detection logic contributed by the community across SIEM and rule languages.
40,407 detections
Filters
Last updated
All Time
Detection languages
15,618
13,637
2,824
1,803
1,727
Contributors
7,678
5,999
5,315
4,504
4,206
Categories
18,287
9,897
3,918
3,796
3,721
Platforms
40,407
7,473
6,580
3,935
3,713
Products / Services
10,635
9,722
6,762
1,902
1,895
MITRE Techniques
13,801
13,234
8,133
6,014
4,458
CVEs
49
45
30
30
29
IDS Classtypes
214
56
43
23
19
IDS Protocols
183
172
20
16
8
IOC sweep for the Optimizely supply chain attack distributing Coruna/DarkSword, covering Defender XDR tables only (DeviceNetworkEvents, DeviceEvents and DeviceFileEvents). Matches remote URLs and hosts, DNS query responses, and IOC IPs against the attacker domains and IPs, and matches file SHA256 hashes against the known malicious samples. Companion to the Sentinel lake/auxiliary tables query, which covers CommonSecurityLog and DnsEvents. Keeps a 14-day lookback in the query body.
Detects OUTLOOK.EXE initiating an outbound SMB connection (TCP ports 445 or 139) to a public IP address. This behavior is consistent with exploitation of Microsoft Outlook NTLM hash leak vulnerabilities where a crafted message forces Outlook to authenticate to an attacker-controlled SMB server, potentially leaking the user's Net-NTLM hash.
Detects suspicious commands as identified in CISA Living off the Land pdf. Alone they may be normal but in concert, they may be worth looking into
Native curl launched from cmd or a web process to download PowerShell/batch scripts, with scripts generated and staged on Windows database servers.
Detects when a Java process establishes an outbound TLS connection (port 443 or 8443) to a public IP address, and subsequently spawns a shell or scripting child process within 10 minutes, filtering out known benign baseline parent-child command line pairs.
hunts ARTEX agents trying to turn database access into code execution by abusing MySQL logging to place JSP web shells in web-accessible directories, with compile-and-retry cycles until output returned. A database engine has no legitimate reason to create .jsp/.php/.aspx files.
Pivot: Expect multiple rewrites of the same file, since agents self-correct after compile errors.
Pivot: Expect multiple rewrites of the same file, since agents self-correct after compile errors.
Detects rare command line executions where cmd.exe launched from explorer.exe executes both an MSI package and an Excel spreadsheet from a mounted image, a pattern associated with BITTER threat actor activity.
An IOC sweep for scoping, not alerting. It matches the relay domain and IP, the wave artifacts (wupdate_<8hex>.msi, wclient_<8hex>), and the rogue client hash where the hash is populated.
Hunt: Lists processes started directly by a JVM that show attacker tradecraft: reverse shells, download-and-execute or decode-and-execute, execution from a staging directory, persistence, or secrets harvesting. Recon tools count only as a burst of 3 or more different tools from one JVM, the typical RCE-validation pattern (whoami, id, uname, hostname). One row per parent JVM, with one evidence line per child showing signal, executable path and full command line.
Detects reflective PE loading inside a single process (RWX allocation, then PE/MZ write, then memory protection change within 10 minutes), as used by BubbleLoader's third stage in hijacked host applications. Baselined with a 30-day process-hash prevalence check (rare binaries only) and JIT/runtime exclusions.
Detects msbuild.exe running an inline CodeTaskFactory task that connects to a public address and then launches a winsvc_*.exe binary from a Temp path, as seen in UAT-12643 (Overlord RAT) campaigns. Excludes common package/Microsoft endpoints and summarizes per device and child process.
Detects ScreenConnect.WindowsClient.exe spawning command-line interpreters (cmd.exe, powershell.exe, wscript.exe, msiexec.exe) pointing to payloads or execution artifacts inside the ScreenConnect client Temp folder (...
ode
ame emp eports or similar), excluding common baseline tooling and self-maintenance tasks.
ode
ame emp eports or similar), excluding common baseline tooling and self-maintenance tasks.
In the CVE-2026-84869 ScreenConnect exploitation chain, the attacker’s script reconfigures the Secondary Logon service (seclogon) to start automatically, then starts it. This lets runas /trustlevel launch the attacker’s rogue ScreenConnect client under a different trust level. This query flags sc.exe setting seclogon to automatic or delayed-automatic start. That change has little legitimate reason on most endpoints. The query only sees sc.exe, so it won’t catch other tools making the same change.
Detects process creation activity characteristic of the Impacket wmiexec tool, where WmiPrvSE.exe spawns cmd.exe executing commands with output redirection to the ADMIN$ share. Enriched with preceding network logons for context.
Detects OUTLOOK.EXE connecting to public IP addresses over SMB (445/139), which Outlook does not normally do. May indicate forced NTLM authentication and NTLMv2 hash leakage via a crafted message file (CVE-2026-124194, exploited in the wild). The exact file extension of the trigger is not confirmed; likely candidates are .msg and possibly other Outlook message formats (.eml, .oft).
The following analytic detects a non-Chrome process accessing files in the Chrome user default folder. It leverages Windows Security Event logs, specifically event code 4663, to identify unauthorized access attempts. This activity is significant because the Chrome default folder contains sensitive user data such as login credentials, browsing history, and cookies. If confirmed malicious, this behavior could indicate an attempt to exfiltrate sensitive information, often associated with RATs, trojans, and advanced persistent threats like FIN7. Such access could lead to data theft and further compromise of the affected system.
The following analytic detects the execution of the `Get-DomainGroupMember` cmdlet from PowerView, identified through PowerShell Script Block Logging (EventCode=4104). This cmdlet is used to enumerate members of elevated domain groups such as Domain Admins and Enterprise Admins. Monitoring this activity is crucial as it indicates potential reconnaissance efforts by adversaries to identify high-privileged users within the domain. If confirmed malicious, this activity could lead to targeted attacks on privileged accounts, facilitating further compromise and lateral movement within the network.
runas /trustlevel:0x20000 starts the payload from a user-writable path (the rogue client in %LOCALAPPDATA%). It follows the msiexec /a step and relies on the Secondary Logon service the actor enabled.
The actor unpacks the rogue ScreenConnect client with msiexec /a (administrative image) into %LOCALAPPDATA%\wclient_<8hex>. This leaves no install record. The query keys on the behavior, not the folder name, so it survives prefix changes.
Catches the path-traversal persistence attempt (..\..\..\Startup\Sync.bat). The host client has no legitimate reason to touch a Startup folder. Matches both the long path and the 8.3 short name (STARTM~1) from the traversal string.
Hunting query for a CoffeeLoader variant delivered by a revived Retadup USB worm. The loader drops a patched 7-Zip DLL into a folder under ProgramData or AppData\Local and names both the folder and the file with an eight hex digit value. It starts the DLL through rundll32.exe from a Run key value. The query returns three signals per device. The first is a Run or RunOnce value that starts such a DLL through rundll32.exe. The second is a rundll32.exe process that loads such a DLL. The third is a connection to a command and control or hosting domain published for the campaign, including its subdomains. Logic tested with synthetic events. Not yet validated against live infections.
Process injection into dllhost, WerFault, svchost, OpenWith and AtBroker is not covered. The USB spread, the DLL write itself and DNS lookups without a network connection record are not covered. The published request path (/payload?md5=) is not covered because network events usually carry only the host name. Connections recorded by IP only are not covered, and the research published no IP indicators. Run values written before the lookback window or before a device was onboarded produce no event. Other autorun locations (RunOnceEx, RunServices, Policies\Explorer\Run) and renamed copies of rundll32.exe are not covered. The domain list only holds what the original research published.
Process injection into dllhost, WerFault, svchost, OpenWith and AtBroker is not covered. The USB spread, the DLL write itself and DNS lookups without a network connection record are not covered. The published request path (/payload?md5=) is not covered because network events usually carry only the host name. Connections recorded by IP only are not covered, and the research published no IP indicators. Run values written before the lookback window or before a device was onboarded produce no event. Other autorun locations (RunOnceEx, RunServices, Policies\Explorer\Run) and renamed copies of rundll32.exe are not covered. The domain list only holds what the original research published.
Page 3 of 1925