WatchGuard H1 2026 Report Finds Highly Targeted Malware and Persistent Attacks on Old Flaws

Summary
WatchGuard’s H1 2026 report says 95.72% of Q2 endpoint malware detections appeared on one machine, while network attacks targeted old flaws. It also found 95% of detected network malware used TLS, which only 20% of Fireboxes inspect.
Key points
- The report draws on opted-in telemetry from tens of thousands of Fireboxes and detections across WatchGuard-protected endpoints.
- In Q2, 95.72% of endpoint malware alerts appeared on one machine; never-before-seen endpoint malware rose 2,065% year over year.
- Network attack detections per Firebox fell about 79%, while unique IPS signatures increased from 381 to 410.
- The median vulnerability referenced among the top 50 IPS signatures was disclosed in 2014; 31 of 44 CVE-referenced signatures involved flaws at least a decade old.
- WatchGuard reports that 95% of network malware arrived over TLS, although only 20% of deployed Fireboxes inspect encrypted traffic. On devices with IAV and APT Blocker, 28% of detected malware evaded signature-based methods.
- Behavioral and machine-learning engines accounted for over 67% of Q2 endpoint detections. Credential access led Q1 threat-hunting tactics, and process hollowing made up 77.1% of Q2 exploit-technique alerts.
- The report recommends prioritizing patches for old, actively targeted flaws, retiring unsupported edge devices, improving TLS inspection, and strengthening Linux monitoring.
Article Details
- Publisher
- WatchGuard
- Report Period
- H1 2026
- Scope
- Global network and endpoint threat detections and public extortion claims.
- Sample Size
- Tens of thousands of opt-in Fireboxes; network attack regional analysis covered 3,858 IPS-enabled Fireboxes. Endpoint telemetry covered millions of protected endpoints.
- Key Statistics
- In Q2 2026, 95.72% of endpoint malware alerts involved a file or process observed on exactly one machine.
- In H1 2026, 95% of detected network malware arrived over TLS, while 20% of deployed Fireboxes inspected encrypted traffic.
- Average network attack detections fell from 845 per Firebox in H2 2025 to 181 in H1 2026, while the number of unique IPS signatures firing rose from 381 to 410.
- Of the top 50 network attack signatures, 44 referenced specific CVEs; 31 of those targeted flaws at least a decade old, and the median disclosure year was 2014.
- WatchGuard tracked 4,932 public extortion claims in H1 2026; the top eight groups accounted for 50.79% of claims, and 41 new groups appeared.
- Recommendations
- Enable TLS inspection where feasible and pair it with behavioral and machine-learning malware detection.
- Prioritize patching based on exploitation evidence, inventory internet-facing legacy applications, and retire unsupported edge and security appliances.
- Monitor Linux preload configuration and system calls, segment IoT devices, and investigate unexpected proxy traffic or CPU spikes.
- Disable unused HTTP methods, monitor web-accessible directories for new executable files, and use parameterized queries and input validation.
- Track attack reach as well as volume and enable layered IPS protections across network segments.
Indicators of compromise
| Type | Indicator | Context |
|---|---|---|
| MD5 | 213eca72f00563fa2ed788a1212c67e0 | Malicious sample attributed to the TamperedChef campaign in Q2 2026. |
| MD5 | 2413b273c989c846441e9bcebbe0daba | GuLoader sample detected on endpoints in Q1 2026. |
| MD5 | 2b313ceea938698985abf3cb5daa8d7e | Unidentified malicious dropper detected in Q1 2026. |
| MD5 | 5261c38bfc6784e95285ad21e62f82fd | Malicious sample attributed to the TamperedChef campaign in Q1 2026. |
| MD5 | 7d9542ef7c46ed5e80c23153dd5319f2 | Conficker Worm sample detected in both Q1 and Q2 2026. |
| MD5 | a34ed3654424f72b99ff6da694d7388e | Malicious sample attributed to the TamperedChef campaign in Q1 2026. |
| MD5 | a4f2c851f5d4336faa3f0955f9008326 | Coin miner dropper sample detected in Q1 2026. |
| MD5 | c379aff38f5f05429158a520d896a136 | Gh0stRAT sample detected in Q1 2026. |
| MD5 | c38f92b1484e0ffeb3c30402d7a6beac | Floxif file-infector sample detected in Q2 2026. |
| MD5 | e2a2521cb16da1bed01565c503772125 | Conficker Worm sample detected in Q2 2026. |
| MD5 | f46329e59f449cdcd96a1d78b4e96f59 | GuLoader sample detected on endpoints in Q1 2026. |
| MD5 | f6bc17c61524935637887617e5752f88 | Malicious sample attributed to the TamperedChef campaign in Q2 2026. |
| URL | hxxps[:]//cloud-onedriveid-bldc23amp[.]meetliaross[.]info/ | Phishing URL embedded in a malicious OneDrive-themed PDF; reported inactive at the time of analysis. |
MITRE ATT&CK
T1053.005 · Scheduled TaskTamperedChef used schtasks.exe for scheduled-task persistence.T1055.012 · Process HollowingProcess hollowing accounted for 77.1% of Q2 exploit technique alerts.T1505.003 · Web ShellA generic web shell command signature was the most widespread network attack detection.T1547.001 · Registry Run Keys / Startup FolderQ2 persistence detections included registry AutoStart keys and startup folders.T1555.003 · Credentials from Web BrowsersQ1 threat hunting alerts included credential theft from web browsers.T1562.001 · Disable or Modify ToolsEndpoint exploit alerts included attempts to bypass AMSI.T1574.006 · Dynamic Linker HijackingThe analyzed libprocesshider rootkit uses LD_PRELOAD to load a library that hides processes.
CVE
CVE-2014-2850Sophos · Web appliance; Sophos · Web appliance firmwareApache Struts remote code execution, and a command injection in a discontinued Sophos Web Appliance (CVE-2014-2850).CVE-2016-3081Apache · Struts; Huawei · Agile controller-campus firmware; +18 moreagainst a family of remote code execution vulnerabilities in the Apache Struts 2 framework from 2016 (CVE-2016-3081 and related), with detection logic that also covers the infamous CVE-2017-5638, the Jakarta MultipartCVE-2017-11882Microsoft Corporation · Microsoft OfficeFinally, `Exploit.CVE-2017-11882.Gen` (52,782 detections) shows that older Microsoft Office equation-editor exploits continue to find victims.CVE-2017-12615Apache · Tomcat; Netapp · 7-mode transition tool; +20 moretargeted flaws between nine and twenty-three years old. These include Apache Tomcat HTTP PUT (CVE-2017-12615), Apache Struts remote code execution, and a command injection in a discontinued Sophos Web ApplianceCVE-2017-5638Apache Software Foundation · Apache Struts2 framework from 2016 (CVE-2016-3081 and related), with detection logic that also covers the infamous CVE-2017-5638, the Jakarta Multipart parser flaw exploited in the 2017 Equifax breach. Struts vulnerabilities haveCVE-2020-6754Dotcms · Dotcms| 1136822 | Web threats | WEB dotCMS CMSFilter assets Access Control Weakness (CVE-2020-6754) | Network Device, Others | 10.96% |CVE-2021-44228Apache Software Foundation · Apache Log4j2| 1230275 | Web threats | WEB Apache log4j Remote Code Execution -2.h (CVE-2021-44228) | Windows, Linux, Freebsd, Other Unix | 3.72% |CVE-2023-25725Debian · Debian linux; Haproxy · HaproxyThe SHELLCODE NOP Sled signature dropped from #3 to #24, and the HAProxy header bypass (CVE-2023-25725) fell from #2 to #36 by volume.
Threat Actors
QilinNamed as one of the two groups associated with 23.30% of WatchGuard's tracked public extortion claims in H1 2026.TeamPCPWatchGuard reports that the group compromised GitHub repositories and stole data from organizations during Q1 supply-chain attacks.TeamTNTThe report says earlier TeamTNT activity reused process-hiding code to conceal mining processes after compromise.The GentlemenNamed as one of the two groups associated with 23.30% of WatchGuard's tracked public extortion claims in H1 2026.
Malware
Agent Tesla| | | | | Mimikatz is an open-source post exploitation hacking tool | | Agent Tesla is another information stealer and remote access | |Application.Agent.IIQdetections and a related variant, `Heur.BZC.PZQ.Boxter.919`, at 409,801. The #2 family overall, `Application.Agent.IIQ` (605,127 detections), is also a dropper and arrived entirely over encrypted connections. A thirdConfickerretire.** GuLoader was the most prevalent endpoint malware in Q1, joined by coinminers, TamperedChef, Conficker (2008), and Gh0stRAT, whose source code leaked in 2008. Mimikatz and Floxif appeared in Q2, and AutoKMSFloxifcoinminers, TamperedChef, Conficker (2008), and Gh0stRAT, whose source code leaked in 2008. Mimikatz and Floxif appeared in Q2, and AutoKMS licensing-bypass tools made up half of the Q2 top 10 PUPs.Gh0stRATwas the most prevalent endpoint malware in Q1, joined by coinminers, TamperedChef, Conficker (2008), and Gh0stRAT, whose source code leaked in 2008. Mimikatz and Floxif appeared in Q2, and AutoKMS licensing-bypassGuLoader- **Old malware refuses to retire.** GuLoader was the most prevalent endpoint malware in Q1, joined by coinminers, TamperedChef, Conficker (2008), and Gh0stRAT, whose source code leaked in 2008. Mimikatz and FloxifHeur.BZC.PZQ.Boxterof the network malware list.** Four of the top 10 families by volume were droppers, led by `CMD:Heur.BZC.PZQ.Boxter` at 710,699 detections and a related variant, `Heur.BZC.PZQ.Boxter.919`, at 409,801. The #2 familyHeur.Mint.Zard.24`Heur.Mint.Zard.24`, a family that has resurfaced in prior periods, appears as a botnet with 96,602 detections.Mirairesidential proxy client posted 184,522. Two Linux coinminer families combined for another 214,801 and a Mirai variant was our most widespread malware detection, peaking at 28.75% of Fireboxes in Sweden.Trojan.Valyria.8923`Trojan.Valyria.8923` contributes another 113,318 detections as a dropper.
Vendors
Products
AD360malware blocked on the millions of endpoints protected by WatchGuard’s Endpoint Security portfolio and AD360. Total threats held roughly steady while never-before-seen threats kept climbing, nearly every alertApache Tomcatfive first-time top-50 entrants targeted flaws between nine and twenty-three years old. These include Apache Tomcat HTTP PUT (CVE-2017-12615), Apache Struts remote code execution, and a command injection in aAPT Blocker- Average malware detections by APT Blocker per Firebox: 61 (35% decrease HoH)dotCMSThe signatures that dominated late 2025 collapsed the hardest: detections for the dotCMS access control weakness that had led our list since mid-2025 fell 89%, and the HAProxy header-bypass exploit that ranked #2 lastEndpoint Securitythe Firebox, we inspect the malware blocked on the millions of endpoints protected by WatchGuard’s Endpoint Security portfolio and AD360. Total threats held roughly steady while never-before-seen threats keptFirebox- [Firebox Feed](#what-is-the-firebox-feed)Gateway AntiVirusSignature-based Gateway AntiVirus hits fell 43%.HAProxydetections for the dotCMS access control weakness that had led our list since mid-2025 fell 89%, and the HAProxy header-bypass exploit that ranked #2 last period fell over 99%, dropping from more than 322,000IntelligentAVProactive IntelligentAV detections rose 49% while its signature-based sibling fell.Intrusion Prevention ServiceThe Firebox’s Intrusion Prevention Service (IPS) blocks known software exploits aimed at servers and clients. Volume collapsed to roughly one blocked attack per Firebox per day, down from nearly five, while the varietySophos Web Appliance(CVE-2017-12615), Apache Struts remote code execution, and a command injection in a discontinued Sophos Web Appliance (CVE-2014-2850).
Tools
libprocesshider- **Linux, IoT, and proxy abuse featured strongly.** A rootkit matching the opensource `libprocesshider` tool posted 201,808 detections while the IPRoyal `Pawns.app` residential proxy client posted 184,522. Two LinuxMimikatzjoined by coinminers, TamperedChef, Conficker (2008), and Gh0stRAT, whose source code leaked in 2008. Mimikatz and Floxif appeared in Q2, and AutoKMS licensing-bypass tools made up half of the Q2 top 10 PUPs.XMRigIf the process name matches a hardcoded target (for example “xmrig” or a custom payload name), that PID directory is silently dropped from the results.
Countries
AustraliaIt shows the highest concentration in Sweden (28.75%), followed by the Netherlands (16.67%) and Australia (14.21%).Belgiumdetected on 52% of Fireboxes in the Americas and 42% in EMEA, including 75% of reporting Fireboxes in Belgium, 60% in Italy, and 59% in the United States. In Belgium, that saturation came from just 774 detectionsGermany`Trojan.MSIL.Basic.8.Gen` follows with significant activity in Hong Kong, Germany, and Turkey.India`Trojan.Generic.38227153` shows the most concentrated impact in a single country, with India accounting for 40.67% of its detections among the top three.Italy52% of Fireboxes in the Americas and 42% in EMEA, including 75% of reporting Fireboxes in Belgium, 60% in Italy, and 59% in the United States. In Belgium, that saturation came from just 774 detections spread acrossNetherlandsIt shows the highest concentration in Sweden (28.75%), followed by the Netherlands (16.67%) and Australia (14.21%).Sweden214,801 and a Mirai variant was our most widespread malware detection, peaking at 28.75% of Fireboxes in Sweden.United Statesand 42% in EMEA, including 75% of reporting Fireboxes in Belgium, 60% in Italy, and 59% in the United States. In Belgium, that saturation came from just 774 detections spread across three-quarters of the
