A blocked C2 domain does not necessarily stop malware from finding its next server. A dead drop resolver gives it another place to look. When that lookup data sits on a public blockchain, removing it is much harder than taking down a website.
On 17 September, Chainalysis reported a 420% increase in blockchain dead drop activity over twelve months. Its cases include North Korean operators, suspected Iranian state-linked actors and Russian-language criminals. The technique is older than those campaigns: the report traces examples back to Namecoin in 2013.
You don't need to run a cryptocurrency business to encounter this. If a process retrieves instructions from an allowed service and then connects to an unfamiliar server, can your team trace what happened? That is worth checking in your own environment.
What a dead drop resolver does
Instead of carrying a fixed command-and-control (C2) address, malware retrieves the current address from an external location. MITRE calls this T1102.001: Dead Drop Resolver. The attacker may control the content or account used for the lookup without controlling the underlying platform.
Traditional examples use paste sites or social media profiles. Blockchain variants store pointers, configuration or malware payloads in transaction data or smart contracts. A lookup may return a server address or executable code, depending on the stage of the infection.
In these cases, the blockchain helps malware find or load the next stage. Commands and stolen data then pass through servers outside the blockchain.
For the BNB Smart Chain (BSC) contract examined by Google Threat Intelligence Group (GTIG), more than 20 updates in the first four months cost an average of $1.37 each at the time.
Three reported variants
DPRK: two linked techniques
Chainalysis connected two techniques to UNC5342, the North Korean threat group tracked by GTIG. In the transaction-based variant, malware first queries Tron for a pointer. If that fails, it tries Aptos. Both routes point to a BSC transaction containing encrypted instructions, including C2 addresses. Once decoded, those instructions lead to off-chain infrastructure.
GTIG's October 2025 report describes a different chain. An initial downloader retrieves JADESNOW from a BSC smart contract through a centralised API provider. JADESNOW then retrieves payloads from Ethereum transaction data, including the INVISIBLEFERRET.JAVASCRIPT backdoor and an additional credential-stealing component. The campaign uses fake recruitment exercises and malicious packages to reach developers.
Chainalysis linked this transaction-based variant to the BSC deployer in GTIG's investigation. The reports describe two techniques sharing infrastructure.
Two linked techniques
UNC5342 · Separate execution paths
BSC / Ethereum payload delivery
GTIG · October 2025
- Step 1Initial downloader
Retrieves JADESNOW through an API service.
- Step 2BSC contract
Hosts JADESNOW; the loader retrieves the next stage.
- Step 3Ethereum records
Backdoor and additional payload components.
- Step 4Off-chain C2
Backdoor communicates with the attacker’s server.
Transaction-based C2 lookup
Chainalysis · September 2026
- Step 1Infected device
Queries a pointer; tries the fallback if needed.
- Step 2Tron or Aptos
Tron is the primary lookup. Aptos is the fallback.
- Step 3BSC transaction
Encrypted instructions include C2 addresses.
- Step 4Off-chain C2
Malware connects to the resolved destination.
Shared deployer infrastructure links the operations. It does not establish a single four-chain execution path. Retrieval services remain intervention points.
Suspected Iran-linked activity: Bitcoin transaction data
Chainalysis describes operators suspected of links to Iran's Ministry of Intelligence placing C2 routing data in Bitcoin OP_RETURN fields. Small payments to a well-known Satoshi-associated address give the malware a stable place to search for those transactions. Using the address as a lookup reference does not implicate its owner.
The attribution comes from malware and operational evidence, including decoding logic and C2 infrastructure. Blockchain records alone do not establish the Iranian link.
Russian-language criminals: shared resolver infrastructure
The report also describes a subscription-based malware operation using Polygon smart contracts to hold infrastructure pointers. An operator updates the contracts for downstream campaigns. The Russian-language tooling and marketing do not establish state sponsorship.
Chainalysis connects the deployer to a wider footprint, including fraudulent tokens, clipboard-hijacking campaigns and more than 50 similar resolver contracts on BNB Smart Chain. The wider footprint therefore extends beyond Polygon.
Three reported cases
Storage, retrieval and defender intervention
DPRK / UNC5342
Two linked techniques
- Storage
- BSC contract and Ethereum data; Tron or Aptos pointers to a BSC transaction.
- Retrieval
- GTIG: centralised API services. Chainalysis: primary and fallback lookups.
- Intervention
- API-provider action, endpoint isolation and off-chain C2 disruption.
- Evidence limit
- Shared deployer links activity, not one combined execution path.
Suspected Iran link
Bitcoin OP_RETURN
- Storage
- C2 routing data in transactions, with a stable recipient as the lookup point.
- Retrieval
- Malware finds and decodes the relevant transaction data.
- Intervention
- Control retrieval access, isolate endpoints and disrupt off-chain C2.
- Evidence limit
- Attribution relies on malware and wider operational evidence.
Russian-language crime
Polygon resolver service
- Storage
- Polygon contract pointers. Wider footprint includes 50+ BSC contracts.
- Retrieval
- Malware reads the contract and follows its current pointer.
- Intervention
- Control retrieval access, isolate endpoints and disrupt off-chain C2.
- Evidence limit
- Language is not proof of state sponsorship.
Preserving a blockchain record does not guarantee malware can still retrieve or use it.
Published indicators
GTIG identifies the BSC contract 0x8eac3198…ad48a71c and its owner, 0x9bc1355344b54dedf3e44296916ed15653844509. Its report also identifies these Ethereum transactions:
0x86d1a21f…a33a: an INVISIBLEFERRET.JAVASCRIPT payload.0xc2da361c…cd0f: a split JavaScript payload that runs the backdoor and installs a Python interpreter for an additional component.0xf9d43274…ce41: the additional credential-stealing component.
GTIG analysed the payloads and published these indicators. The explorer links show the underlying transaction records; the report explains what the code does.
These records remain useful for investigating known activity. Attackers can still move to new transactions, contracts or wallets, so detections tied to these addresses alone will miss that change.
Where defenders can intervene
GTIG reports that some API providers acted when notified about UNC5342's activity. Blocking retrieval through those services can disrupt the malware even while the underlying blockchain records remain available.
Restrict outbound access where there is no business need for it. A cryptocurrency business will need exceptions that other organisations may not. Blocking known providers still leaves gaps if malware can use another service or attacker-controlled infrastructure.
You can also isolate the endpoint or disrupt the off-chain C2 server. A fallback lookup helps malware recover from a failed request; it does not prevent either of those actions.
Check what your logs capture
Start with the process making the request. A developer tool or wallet querying a blockchain API may be expected. A script interpreter launched from a downloaded interview project deserves different scrutiny. Record which process launched it and where it connected next.
DNS and connection logs can identify destinations, but they do not normally reveal an encrypted HTTPS request's method or body. Detecting a particular contract address or an eth_call request needs telemetry that exposes that detail, such as suitable application logging or authorised TLS inspection. Do not assume an EDR's network events contain it.
Read-only calls do not create a blockchain transaction. A public explorer can show an operator's writes, not a list of your infected devices reading them. To find those reads, you need endpoint, proxy or provider-side records.
What to test, and what each result proves
Use benign data and destinations controlled for the exercise. There is no need to retrieve live malware or contact an attacker's server. Agree the process, device groups, allowed destinations and containment actions before the test. A testnet can be useful, but its providers and reputation may differ from the production services in the reported campaigns.
- Retrieval visibility. Make an approved read-only request from the selected test process. Check whether it is allowed or blocked and whether the monitoring records the process, destination and timestamp. This establishes access and visibility only. A successful request without an alert is not automatically a control failure.
- Follow-on connection. Have the benign test resolve an exercise-controlled destination and connect to it. Check whether an analyst can connect the retrieval event to the later connection, including process ancestry if different processes perform the two steps. Save the event identifiers and the analyst's reasoning.
- Changed destination. Repeat with a second approved destination absent from the exercise's initial indicator list. Check whether the correlation still works. This distinguishes behaviour-based coverage from a rule that only recognises the first hostname.
- Containment. With approval, apply the agreed endpoint or network control and repeat the benign sequence. Verify that follow-on communication is prevented. Record time to triage and containment, and identify the evidence that initiated the response.
Where blockchain use is expected, include a legitimate workload for comparison. If both activities alert, what lets the analyst tell them apart? If neither alerts, first check whether the required events were collected.
In the purple team exercise report, record which devices and logs were tested, what failed and who owns the fix. A missing event needs a different fix from an alert that nobody acted on. Retest the failed stage once the change is in place.
About the evidence
The campaign findings and attribution are from Chainalysis and GTIG. The 420% increase applies to activity Chainalysis tracked; it is not an estimate of an individual organisation's exposure. The testing section proposes checks based on those reports.


