<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Will Archer — malware &amp; threat research</title><link>https://fullchecksecurity.com/</link><description>Recent content on Will Archer — malware &amp; threat research</description><generator>Hugo</generator><language>en-au</language><lastBuildDate>Sun, 22 Nov 2026 00:00:00 +0000</lastBuildDate><atom:link href="https://fullchecksecurity.com/index.xml" rel="self" type="application/rss+xml"/><item><title>A 2014 CosmicDuke-style stealer in a Chrome Updater costume: three transports, one build ID</title><link>https://fullchecksecurity.com/posts/cosmicduke-style-stealer-2014-build/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/cosmicduke-style-stealer-2014-build/</guid><description>A static look at a 2012-era loader wrapping a 2014 stealer whose config lists FTP, WebDAV and HTTP channels for both upload and tasking. Includes the wire formats, the build markers, and why a sandbox tagged it with nine unrelated malware families.</description></item><item><title>A registrar, a password, and node -e: a BeaverTail stage-4 with PBKDF2 and AES-256-CBC</title><link>https://fullchecksecurity.com/posts/beavertail-stage4-registrar/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/beavertail-stage4-registrar/</guid><description>Static RE of a 7,867-byte BeaverTail stage-4 loader. Two unauthenticated bare GETs: a /s/&lt;!-- raw HTML omitted --&gt;/&lt;!-- raw HTML omitted --&gt; registrar whose response starts with the ASCII prefix &amp;lsquo;JS6&amp;rsquo; and base64-decodes to a CSV naming the /main host, and a /main fetch returning salt:iv:ct hex that decrypts to JavaScript under a hardcoded password via PBKDF2-SHA256/100k. The next stage then runs under &lt;code&gt;node -e&lt;/code&gt;. Password recovered. No execution, no live fetch.</description></item><item><title>A ShadowPad test build: the packet codec, and what the "MyTest" markers tell you</title><link>https://fullchecksecurity.com/posts/shadowpad-test-build-packet-codec/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/shadowpad-test-build-packet-codec/</guid><description>A static walk through one ShadowPad DLL: the stage-1 decryption, the 16 plugins in Root, the 20-byte packet header and its cipher, and why six copies of the string MyTest make me think this is a developer or tester build rather than an operational one.</description></item><item><title>An NPS fork with teeth: static RE of a SNOWLIGHT / VShell stage-2</title><link>https://fullchecksecurity.com/posts/snowlight-vshell-stage2/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/snowlight-vshell-stage2/</guid><description>A stripped, garble-obfuscated Go ELF delivered by the SNOWLIGHT / UNC5174 stager chain. Static RE recovers the config (AES-128-CBC from MD5(vkey)), the AES-256-GCM record layer and NPS-derived stream multiplexer, the full command set including LocalRun&amp;rsquo;s in-process shellcode execution and PluginRun&amp;rsquo;s memfd+LD_PRELOAD trick, and the kill-switch files. No detonation, no C2 contact.</description></item><item><title>nginx eats the NULs: recovering PolinRider stage 3 from a detonation pcap</title><link>https://fullchecksecurity.com/posts/polinrider-stage3-nulstrip/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/polinrider-stage3-nulstrip/</guid><description>PolinRider stage-3 (/$/boot) couldn&amp;rsquo;t be decoded by anyone, including by the operator&amp;rsquo;s own next-stage. It isn&amp;rsquo;t fingerprint-gated — the operator&amp;rsquo;s nginx/PHP layer strips every 0x00 byte from the body, so about 0.9% of a repeating-key XOR stream vanishes and the decode falls apart at the first plaintext char == key char (offset 144). Here&amp;rsquo;s the mechanism, the reconstruction (n-gram beam search over &amp;lsquo;byte-was-dropped-here&amp;rsquo; + RC4 key-pool brute force on the string table), and what the recovered stage does: sandbox blocklist, environment exfiltration via /snv, injection into developer tooling (VS Code, Cursor, Antigravity, Discord, GitHub Desktop, global npm cli.js), a clipboard stealer targeting LLM-API keys alongside wallet material, and the operator&amp;rsquo;s own test-machine hostnames in the strings.</description></item><item><title>One ASN, nine live C2s in a week: what a reseller network's prefix space looks like from ThreatFox</title><link>https://fullchecksecurity.com/posts/one-asn-nine-c2s-intrusion-hosting/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/one-asn-nine-c2s-intrusion-hosting/</guid><description>A one-week ThreatFox sweep across all 68 IPv4 prefixes of AS399629, which families showed up and which didn&amp;rsquo;t, and a licence-file fingerprint that links one NetSupport RAT operator across two hosters. About the network and its tenants, not the people behind it.</description></item><item><title>Three TLS layers and two yamux sessions: reading a BRICKSTORM backdoor without running it</title><link>https://fullchecksecurity.com/posts/brickstorm-wire-shape-static-re/</link><pubDate>Sun, 22 Nov 2026 00:00:00 +0000</pubDate><guid>https://fullchecksecurity.com/posts/brickstorm-wire-shape-static-re/</guid><description>A static teardown of one BRICKSTORM Go sample: how it hides its C2 behind DNS-over-HTTPS, WebSocket and three nested TLS handshakes, what the four command bytes do, and the network signatures that fall out of it. No execution, no C2 contact.</description></item></channel></rss>