Skip to content
NewCybercrime Report 2026 · Eight Chapters · A Post Is Not An Incident → Explore The ChaptersCybercrime Report 2026 → Explore

Odin Research / What is / Threat groups and feed labels

What Is DarkSword? The iOS Exploit Kit and the C2 Feed Label

One name, two threats: a spyware-grade iPhone exploit chain and a threat-feed label for multi-port C2 servers. Here is how to tell them apart.

By
Odin by Brandefense
Updated
21 Nov 2026
Reading time
7 min read
Type
Feed label

In short

DarkSword is the name of two different things. Google Threat Intelligence Group uses it for a full-chain iOS exploit kit that spyware vendors and suspected state actors used against iPhones on iOS 18.4 to 18.7. A threat feed also uses DarkSword as a family label for multi-port TLS and HTTP C2 servers.

Key takeaways

What it is
Two things share the name: a JavaScript iOS exploit kit documented by GTIG, and a threat-feed family label for multi-port C2 servers.
How it works
The iOS kit chains six flaws through Safari on watering-hole sites. The labelled C2 servers listen on 443, 8888, 2000 and 80, often two ports per host.
Why it matters
GTIG tied the iOS kit to UNC6748, PARS Defense and UNC6353, with targets in Saudi Arabia, Turkey, Malaysia and Ukraine.
What Odin Research saw
Up to 362 DarkSword-labelled endpoints, 28 confirmed. Over half sit on one provider and share a TLS stack with SuperShell, which is weak evidence alone.
What to do first
Update iPhones to the latest iOS and use Lockdown Mode for high-risk users. Block only confirmed DarkSword C2 endpoints.

What is DarkSword?

DarkSword is a name with two meanings in threat intelligence. Google Threat Intelligence Group (GTIG) uses it for a full-chain iOS exploit kit that spyware vendors and suspected state actors aimed at iPhones. Separately, an upstream threat feed uses DarkSword as a family label for multi-port TLS and HTTP command-and-control servers.

This page covers both, because anyone searching the name will meet both. Odin Research reports use the feed label only. They did not check those servers against GTIG's indicators, so the label is not attribution and does not tie the servers to the iOS kit.

A threat feed label names a fingerprint group. It tells you that many servers look alike, not who runs them. Keep the two meanings apart in tickets, block lists and reports.

The DarkSword iOS exploit kit

GTIG published its DarkSword research on 18 March 2026, in coordination with Lookout and iVerify. GTIG named the kit after toolmarks in recovered payloads. It is written entirely in JavaScript, and GTIG says several commercial surveillance vendors and suspected state-sponsored actors have used it since at least November 2025.

  • Targets. iPhones running iOS 18.4 through 18.7.
  • Vulnerabilities. Six flaws in JavaScriptCore, dyld, ANGLE and the XNU kernel. Several were exploited as zero-days. GTIG says all six are fixed by iOS 26.3.
  • Delivery. Watering-hole and decoy websites load the chain through a multi-stage JavaScript loader in Safari. Some variants send visitors who are not targets to legitimate sites. This is a drive-by compromise pattern.

GTIG tied the kit to three actors:

  • UNC6748, which targeted users in Saudi Arabia through a Snapchat-themed decoy site and deployed the GHOSTKNIFE backdoor.
  • PARS Defense, a Turkish commercial surveillance vendor, which targeted users in Turkey in November 2025 and Malaysia in January 2026 with the GHOSTSABER backdoor.
  • UNC6353, a suspected Russian espionage group, which ran watering-hole attacks on Ukrainian websites from at least December 2025 to March 2026. It deployed GHOSTBLADE, a tool that collects messages, keychains, location data, photos and cryptocurrency wallet data.

GTIG reports that all three payloads delete crash logs to reduce forensic evidence. The kit fits the wider mercenary spyware market, and GTIG compares its spread with the earlier Coruna iOS exploit kit.

How the DarkSword C2 feed label works

In Odin Research reports, DarkSword means an upstream threat-feed label for servers that answer on several TLS and HTTP ports. It does not point to one named threat actor or to a known exploit kit. Odin uses it only to group the endpoints the feed tags that way.

The labelled servers share a few habits. Most listen on port 443. Others use 8888, 2000, 80, 3000 or 8443, often two ports on one host. Using a port that does not match its usual protocol is MITRE ATT&CK T1571, Non-Standard Port. HTTPS command channels map to T1071.001, Web Protocols, which lets C2 traffic blend into normal web use.

What Odin Research observed

Three Odin reports cover the DarkSword feed label. Each study draws a different set of records, so the totals differ.

  • One TLS stack, two families. In Monograph 02, 174 of 241 DarkSword records (72.2%), all on CNServer LLC, shared one TLS fingerprint. 126 SuperShell records on a different provider shared the identical fingerprint. That fits a shared C2 bundle just as well as a provider's standard server image.
  • A large, mostly unconfirmed fleet. Monograph 19 counted 346 DarkSword endpoints on 292 hosts, with 10 confirmed. CNServer LLC hosted 189 (54.6%). Nine of the study's 35 multi-listener hubs were DarkSword hubs, and new candidates kept appearing on port 8888.
  • Confirmed servers sit elsewhere. Monograph 28 mapped 362 endpoints on 313 hosts. The 28 confirmed endpoints sat on 19 hosts at providers such as CTG Server Limited (7), Datacamp Limited (4) and YANCY LIMITED (3), not on CNServer. Seven confirmed hosts bound ports 80 and 443 together.
  • High-port pairs. Monograph 28 found 97 high-port listeners, 69 on 8888 and 28 on 2000. Nine candidate hosts exposed both ports.
  • Weak links to other families. Two hosts carried both an ARTEX and a DarkSword listener. Both were unconfirmed, and two hosts out of 2,468 do not show that one group runs both. Monograph 19 also saw DarkSword next to CyberStrikeAI or mitmproxy on some hosts.

The lesson: most DarkSword-labelled servers are leads, not verdicts. A shared TLS fingerprint on a bulk hosting provider is weak evidence on its own.

How to detect DarkSword

For the iOS exploit kit:

  • Check that every managed iPhone runs the latest iOS. GTIG says iOS 26.3 fixes all six flaws.
  • Use the indicators and YARA rules GTIG published for GHOSTKNIFE, GHOSTSABER and GHOSTBLADE in mobile forensic tools.
  • Watch for Safari visits to domains GTIG listed, through DNS or secure web gateway logs.

For the C2 feed label:

  • Alert on non-browser processes that connect to ports 8888, 2000 or 3000 on rare external hosts.
  • Pair any TLS fingerprint match with a second signal, such as a certificate subject or URI. Never block on the fingerprint alone.
  • Treat matches on shared CDN or ingress certificates as noise until proven otherwise. See false positives.

Odin Research publishes rules for the label. Monograph 19 includes a Suricata signature for certificates of three confirmed DarkSword domains on ports 443, 8443 and 8888. Monograph 28 adds Suricata, Sentinel KQL, Sigma and YARA rules for DarkSword check-ins, worker processes and controller configurations. All are in the Defense Kit.

How to defend against DarkSword

  1. Patch iPhones fast. Enforce minimum iOS versions through mobile device management.
  2. Protect high-risk users. GTIG recommends Lockdown Mode for devices that cannot be updated and for users at higher risk of targeted attacks.
  3. Block only confirmed C2. Block the confirmed DarkSword-labelled endpoints from Odin's reports and keep the rest on watchlists. Bulk VPS addresses are reassigned often.
  4. Investigate the process. If a workstation talks to a DarkSword-labelled server, find the process that opened the connection before you call it an infection. Follow how to respond to actively exploited vulnerabilities.

DarkSword compared with similar names and C2 labels

NameWhat it isWhere it shows upNotes
DarkSword (iOS kit)Full-chain iOS exploit kit (GTIG)Watering-hole and decoy websitesUsed by spyware vendors and suspected state actors; fixed by iOS 26.3
DarkSword (feed label)Upstream feed label for multi-port C2 serversPorts 443, 8888, 2000, 80Not an actor; most endpoints unconfirmed
ARTEXAutonomous red-team console familyPort 8787, plus 443 and 80498 confirmed controllers in Monograph 28
SuperShellC2 family label in Odin's feed dataPort 443Shared DarkSword's exact TLS stack in Monograph 02

Frequently asked questions about DarkSword

What is DarkSword?

DarkSword is the name of a full-chain iOS exploit kit documented by Google Threat Intelligence Group in March 2026. The same name is also used by a threat feed as a family label for multi-port C2 servers.

Is the DarkSword C2 label the same as the DarkSword iOS exploit kit?

No link has been shown. Odin Research did not check DarkSword-labelled servers against GTIG's indicators, so the label is not attribution to the iOS kit or its users.

Who uses the DarkSword iOS exploit kit?

GTIG tied it to UNC6748, the Turkish commercial surveillance vendor PARS Defense, and UNC6353, a suspected Russian espionage group.

Which iPhones does DarkSword affect?

GTIG says the kit targets iOS 18.4 through 18.7. All six vulnerabilities it uses are fixed by iOS 26.3, so updating to the latest iOS removes the risk from this chain.

How do I protect my iPhone from DarkSword?

Install the latest iOS update. GTIG also recommends Lockdown Mode for devices that cannot be updated and for people at higher risk of targeted attacks.

Is DarkSword a threat actor?

No. In Odin Research reports, DarkSword is a feed label for a group of similar-looking C2 servers. It names a fingerprint group, not a group of people.

How do I detect DarkSword C2 servers?

Look for non-browser traffic to ports 8888, 2000 and 3000 and pair any TLS fingerprint match with a certificate or URI signal. Odin Research publishes Suricata, KQL, Sigma and YARA rules in its Defense Kit.

How is DarkSword related to Coruna?

GTIG describes Coruna as an earlier iOS exploit kit whose spread resembles DarkSword's.

Sources

  1. Google Threat Intelligence Group, "The Proliferation of DarkSword: iOS Exploit Chain Adopted by Multiple Threat Actors", 18 March 2026. cloud.google.com/blog/topics/threat-intelligence/darksword-ios-exploit-chain (checked 21 November 2026)
  2. MITRE ATT&CK, T1571 Non-Standard Port. attack.mitre.org/techniques/T1571/ (checked 21 November 2026)
  3. MITRE ATT&CK, T1071.001 Application Layer Protocol: Web Protocols. attack.mitre.org/techniques/T1071/001/ (checked 21 November 2026)

Related tools and terms

All Odin Research explanations · Brandefense glossary

Odin Research

Get the rules, not just the definition

Every Odin Research report ships with verified indicators and detection rules you can load today.

Open the Defense Kit