Our own lightweight hypervisor rapidly observes behavior across multiple OS and app versions.
With no agent inside the VM, it resists sandbox detection and sleep-based bypasses.
The Hyper Sandbox runs the document to extract events and behaviors, then chains the behaviors and matches them against attack sequences.
In this sample, the shell → remote download → executable creation → execution sequence matched, and the AI judged it Risk 10 · malicious.
File name, events, scores, techniques and URL: example values from a sample analysis report (analysis flow)
In a decoy Office, macro detection, deobfuscation and vulnerability traps expose hidden malicious code.
It is built on three technologies: Hyper Analysis, Forward Engineering and Morphing Sandbox.
3 TECH
Private Function ylpdfpnmlbmswkm() As Variant
ylpdfpnmlbmswkm = Array("chs", "cht", "deu", "enu", "jpn", "kor")
End Function
Private Function cwqykuktzoyalcwyhbf() As Variant
cwqykuktzoyalcwyhbf = Array("A1", "BU", "CZ", "DA", "HE")
Private Sub dtlqnhxneoffkjywab(kwzbshccypfxpbrc As Boolean)
If Dir(".gattaka") <> "" Then
MsgBox "Not applicable environment", vbCriticalConst merrdjtosyhjjuraz = "16615719117921321028221",
Const iujnoejlhsccsg = "218204209208202280215214"
Dim pnjbsrngsgckjayoasd() As Byte
If Dir(".gattaka") <> "" Thenpnjbsrngsgckjayoasd, False
Set fhfcsnunstz = GetObject(rrhzspxphvvy(...))
dg)).SelectSingleNode("/").Text)In an environment optimized for analysis,
fast and light · solves fragmentation
Forces button and hyperlink clicks
to run for fast detection
An environment that changes every run
neutralizes sandbox evasion
Microservice architecture (MSA) · many small engines come together as one analysis engine.
Splitting the work by role keeps complexity low and makes maintenance and scaling easy.
At run time it reads the file's characteristics and reconfigures the engines in real time for the best result.
Ransomware · infostealers · backdoors · cryptominers and more
Known threats are screened out by the pre-filter first; only the remaining files get deep analysis in the Hyper Sandbox.
AI maps behavior to MITRE ATT&CK techniques, scores the risk and hands it to SOAR · SIEM as report.json.
File creation · deletionProcess creationDLL loadingInjectionRegistry changesC2 connectionsHWPDOCXLSPPTPDFONEVBSJSHTMLHTALNKArchivesZX CORE is a lightweight, fast sandbox analysis engine that handles static, dynamic and AI analysis in one engine.
Its name comes from Zeta(x): analyzing a file (x) with a sandbox (Zeta).
6-layer architecture · under 1 s analysis · 1 million samples a day
Each analysis stage runs on its own, and the results are merged into one.
It is split into six layers: input → static analysis → dynamic analysis → interpretation → output → integration.
File Intake · API GatewayReceives files for analysis and preprocesses themSAMSignature analysis · decryption · deobfuscation · string extractionHyper Sandbox + DBAMRuns files in an isolated environment and monitors behaviorTCE · AI classification engineClassifies and interprets malware with machine learningReporting ModuleGenerates structured analysis reportsAPI · SIEM ConnectorIntegration with external platforms · IOC sharingPlannedFiles received via REST API are statically analyzed, then run in the Hyper Sandbox to collect their behavior.
AI classifies the malware family and threat level, and the result is output as a JSON report.
In the default configuration, 8 instances analyze in parallel and each sample takes under 1 second.
It handles 1 million samples a day, and with an in-house hypervisor and Office, no separate licenses are needed.
Values are based on technical specification v1.4. The 0.27 s file analysis time from the certified test (KBeT) is shown under "Certifications & patents".
VM trace removal · Office vulnerability traps
It runs on our own hypervisor and removes the traces (artifacts) that reveal a virtual environment.
Here is how the environment malware sees differs between a typical VM and ZX CORE.
→ It recognizes the analysis environment and hides
→ It sees a real PC and reveals its malicious behavior
It includes Office vulnerability traps for multiple versions and script emulators.
Documents actually trigger the vulnerabilities they target, and the URLs and C2 IPs they try to reach are extracted.
With traps for multiple Office versions, documents actually trigger the vulnerabilities they target.
Removes obfuscation and extracts behavior command by command.
Extracts the URLs and command server (C2) IPs a file tries to reach while running.
Decompression · decryption · deobfuscation · string decoding
Without running the sample, it analyzes file structure, code and metadata for an initial threat assessment.
It unpacks, decrypts and deobfuscates files, then extracts and decodes encoded strings.
Unpacks archives and analyzes the files inside.
Decrypts encrypted MS Office files (some known methods).
Reveals code hidden with comments and line-joining or line-splitting syntax.
Extracts ASCII and Unicode strings, and automatically decodes strings encoded with Base64, XOR and more.
Static analysis covers 10 document, 3 script, 1 shortcut and 15 archive formats.
DOCDOCXXLSXLSXPPTPPTXPDFRTFONEHWPJSVBSBATLNK7ZBZ2CABGZISOLZ4LZHRARTARTBZXZZIPZSTALZEGG5 anti-evasion techniques that draw out hiding malware
Malware that sleeps for a long time or waits for a click is drawn out within the analysis window.
The five are time acceleration, user action simulation, environment adaptation, multi-execution triggers and random puzzle escape.
Manipulates the internal clock to wake malware that sleeps long before acting, within the analysis window.
Finds and clicks every hidden button and link.
Changes fixed values such as user name, IP, MAC, CPUID and SystemUUID on every run.
Fires multiple events, such as buttons and links, at the same time.
Breaks out of infinite loops that use random values to stall analysis.
0–10 score · 5 levels · 6 families
A machine learning model combines static and dynamic analysis results to classify malware and assign a threat score.
There are five levels in all, from 8–10 points for CRITICAL (malicious) to 0–1 points for BENIGN.
Malware is grouped into six families based on the behavior it shows while running.
Fileless · process execution · user execution lures · auto-run · worm · bot · script execution.
ProcessInjectionShellCodeInjectionetc.DropFileExecDownloadFileExecetc.ExeFileFakerSysFileFakerOleFakeLinkWebFakeExecetc.DropFileAutoRunDownloadFileAutoRunetc.MailDosetc.RemoteCmdExecHiddenPSExecetc.Per-sample results are output as structured JSON for direct integration with automation systems.
An API connects to security operations platforms such as SIEM and SOAR.
Per-sample results are output as structured JSON for direct integration with automation systems.
An API connects to security operations platforms.