Security1 distinct publisher3 min readUpdated
Unit 42 says the C++ loader reads encrypted commands from immutable smart contracts over public RPC endpoints, which turns takedown work into traffic monitoring.
The Watch · Security desk

Compiled by The WatchSomething wrong?How this is made
Palo Alto Networks' Unit 42 has published an analysis of Aeternum, a C++ botnet loader that moves its command-and-control infrastructure entirely onto the public Polygon blockchain, with operators writing encrypted and plaintext instructions directly into smart contracts instead of relying on centralized servers or domains [1][2]. Infected devices then continuously query public remote procedure call (RPC) endpoints to retrieve and execute those on-chain commands [4], which means the first two moves in most response playbooks, seize the server and block the name, have nothing to act on.
The loader itself is unremarkable. The sample Unit 42 examined, named Build.exe, is a UPX-packed 32-bit Windows PE compiled in C++ [10]. It creates a folder under the user's AppData\Local directory, copies itself there, and drops a Startup shortcut following the pattern Wmi_Framework_APIKEY_wmsnet_<random_value>.lnk to survive reboot [12], then executes supporting binaries named wmiframework.exe, ZrvEsJQzWQ.exe and STAAAAAS.exe [13]. It deobfuscates configuration data to build endpoint strings and sends JSON-RPC requests to Polygon RPC endpoints [14].
The part that breaks the playbook is the retrieval step: the loader queries immutable smart contract addresses using the contract method 0xb68d1809 to pull encrypted commands [15]. A smart contract is a self-executing program stored on a blockchain that runs when set conditions are met [3]. Immutable addresses cannot be repointed or edited away, there is no registrar to serve, and reads travel over third-party RPC providers rather than attacker-owned hosts [23]. Unit 42's assessment is that the combination of decentralized C2 and host evasion, including virtual machine detection and antivirus scanning, produces a highly resilient, low-cost threat that complicates existing law enforcement takedown methods [5][6].
Not every leg is beyond reach. The loader downloads files as instructed, including a clean putty.exe and a malicious DotNetZip.dll, from GitHub repositories [17], and that DLL uses hard-coded credentials to reach a Telegram C2 bot [18]. Exfiltration runs over encrypted channels to trusted domains, code-hosting platforms and the Telegram API [19]. Payload delivery and data theft still depend on commercial platforms that answer abuse reports, even though the instruction channel does not [24].
There are also cheap wins for hunters. Unit 42 describes a fixed obfuscation pattern of three null bytes, encrypted payload bytes, a null byte, key bytes, then three null bytes, which lets a script locate every occurrence and its offset and recover the plaintext [20]. Those strings include the JSON objects used for the loader's HTTP-based C2 traffic [22], which is how you get the contract addresses to watch for. The payload decryption itself uses what Unit 42 calls a weak PBKDF2HMAC/AES-GCM routine [16]. Palo Alto Networks says its own products, including Advanced WildFire, Advanced DNS Security and Cortex XDR, cover the threats described [21], which is a vendor claim on a vendor blog.
Watch whether your egress policy has an opinion about JSON-RPC calls to public blockchain endpoints from workstations that have no reason to make them, because that beacon is now the detection surface. Watch for reuse of the same design by other families; the research builds on earlier Ctrl-Alt-Intel work that focused on host-based activity [8], and Unit 42's write-up already links Aeternum activity to Python malware using the Telegram API and to a blended XWorm, XMRig and exfiltration case [7].
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Infected devices continuously query public remote procedure call (RPC) endpoints to retrieve and execute the on-chain commands.
The loader deobfuscates global configuration data to produce parameters used to construct network endpoint strings and sends JSON-RPC requests to Polygon RPC endpoints for decentralized C2 communication.
The loader queries immutable smart contract addresses using the contract method 0xb68d1809 to retrieve encrypted C2 commands.
The loader downloads files as instructed by the C2, such as a clean putty.exe and the malicious DotNetZip.dll, from GitHub repositories.
The malicious DLL uses hard-coded credentials to connect to a Telegram C2 bot (DLLSendC2Bot).
Stolen information is packaged for exfiltration over encrypted channels to trusted domains, code-hosting platforms and the Telegram API.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Detailed first-party artifacts, single publisher
The technical core is unusually concrete for a single post: a named sample with SHA256, packing and architecture, persistence paths and shortcut naming, three supporting binary names, the contract method 0xb68d1809, a reproducible string-decryption pattern, and captured traffic to Polygon RPC, GitHub and the Telegram API. What holds the score down is provenance: every claim traces to one vendor blog, the referenced Ctrl-Alt-Intel prior research is not supplied for corroboration, the supplied body is truncated before the contract-communication analysis completes, and characterizations like 'weak PBKDF2HMAC/AES-GCM' are asserted without parameters.
Confirmed in the wild, scale undisclosed
There is real in-the-wild evidence rather than a lab concept: the malware advertises itself as a loader product, one analyzed sample rotated across 22 smart contract addresses during C2, and three distinct malware cases are tied to the activity, including a Python/Telegram variant and an XWorm plus XMRig blend. But nothing quantifies spread — no victim counts, no geographic or sector distribution, no infection telemetry, no timeline of campaign volume — so the observed footprint supports 'operational' but not 'widespread'.
Resilience framing outruns the article's own caveats
The 'permanent threat / nothing to seize' framing is directionally correct for the instruction-retrieval leg and is well evidenced. It is modestly overstated against details in the same article: payload crypto is described as weak, the string obfuscation is scriptable enough that Unit 42 published the pattern, and both delivery and exfiltration still depend on GitHub, Telegram and other trusted platforms that remain reachable by abuse process. Scale is also undisclosed while the language implies systemic threat, and the piece attaches a vendor product list. The gap is small rather than large because the architectural claim itself holds up.
Vendor research with explicit product attachment
The sole source is a commercial security vendor's research arm, and the article names five Palo Alto Networks product lines as protecting customers plus a paid incident-response contact, with no detection-efficacy data or bypass caveats. That is a direct commercial interest in the threat being read as hard to stop by conventional means. Mitigating factors: prior third-party research is credited by name, and the post publishes reusable IOCs, sample hashes and a deobfuscation method that benefit non-customers too.
Technically solid but single-source and unsized
Confidence in the mechanism is fairly high — the artifacts are specific, internally consistent and checkable against a published hash. Confidence in the significance is much lower: one publisher with a commercial stake, no independent corroboration inside the cluster, a truncated body, and no prevalence or impact figures. The derived takedown and abuse-process conclusions follow logically from the described architecture but are not tested against any law-enforcement or platform-response evidence in the supplied material.
security
Unit 42's Credential Brief: Hunt The Login That Succeeds Right After The Failures1 distinct publisher
security
The EncroChat "national security secret" was exploit code sitting on GitHub1 distinct publisher
build
Pass-ta-key breaks Chrome's device trust, not WebAuthn: harden the endpoint, keep the rollout1 distinct publisher
security
North Korea's hiring funnel: 60 applications a day, 22 personas, ten jobs landed1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.