02 · Multi-layer URL scanning · behavior analysis

URL behavior analysisCatches unknown malicious URLs

From L0 surface analysis to L3 in-depth content analysis, signals build up at each stage to assess risk.
URLs and hyperlinks in email bodies and documents are followed in Hyper Sandbox and judged by behavior.

MULTI-LAYER SCANNING

Multi-layer URL scanning

From a first-pass check with no network requests to in-depth content analysis, signals build up at each stage for a precise risk assessment.
See in the log on the right which scanner checks each layer, from L0 to L3.

  1. LEVEL 0Surface Layer

    Instant analysis from the URL string alone, with no network requests.

    • Surface ScannerStatic URL structure parsing and encoding pattern analysis
  2. LEVEL 1MetaData Layer

    Network metadata inspection layer that sends no request to the target URL.

    • Domain ScannerVerifies domain age and registrant trust; analyzes hosting environment and IP geolocation
  3. LEVEL 2Network Layer

    Network-layer verification after a server connection is established. Inspects HTTP traffic and the certificate chain.

    • Header ScannerDetects anomalies in HTTP headers
    • Cert ScannerChecks trust of the SSL/TLS certificate chain
  4. LEVEL 3Content Analysis Layer

    In-depth content analysis through sandbox rendering. Inspects the DOM tree and script behavior.

    • Html ScannerInspects the DOM tree for suspicious tag patterns
    • JS ScannerJavaScript deobfuscation and behavior analysis
ENGINE_LOGSSTATUS PROCESSING...
  1. [L0]Surface ScannerStatic URL structure parsing and encoding pattern analysisRUN
  2. [L1]Domain ScannerVerifies domain age and registrant trust; analyzes hosting environment and IP geolocation
  3. [L2]Header ScannerDetects anomalies in HTTP headers
  4. [L2]Cert ScannerChecks trust of the SSL/TLS certificate chain
  5. [L3]Html ScannerInspects the DOM tree for suspicious tag patterns
  6. [L3]JS ScannerJavaScript deobfuscation and behavior analysis

Why multiple layers? Filtering threats step by step, from surface analysis at L0 to deep analysis at L3, uses resources efficiently and raises detection accuracy.

URL EXTRACTION · BEHAVIOR ANALYSIS

URL extraction and behavior analysis

We actually follow the URLs behind hyperlinks in a document and analyze the behavior of the pages they lead to before giving a verdict.
In this example PDF, we found a fake CAPTCHA (FakeCaptchaOverlay) on the followed page, analyzed the additional URLs, and judged it Risk 8 · malicious.

  1. Analyzed file · original
    vertexBuffer_nvidia.pdf
    PDF · 184 KB
    • UrlPhishLureT1566.002 · T1204.001
    • HyperLinkHyperlink in document
    • 2 events
  2. URL extraction
    www.nvidia.com/Marketing/De…
    Extracted from a hyperlink in the document
  3. Followed page
    www.nvidia.com.htmRisk 8
    HTM · 4.00 MB
    • FakeCaptchaOverlayT1204.004 · T1189
    • UrlPhishLureT1566.002 · T1204.001
    • 126 events
  4. Additional URLs · pages
    assets.adobedtm.com
    → www.nvidia.com_2.htm · HTM 4.10 MB
  5. AI verdict
    8Risk
    FakeCaptchaOverlayMalicious
    5 targets · 128 events · 2 behaviors · 2 engines · 18.69 s

File names · URLs · risk · events · techniques · time: example based on a sample analysis report (analysis flow)

FORCED EXECUTION

Even links waiting for a click are forced to run

Buttons, hyperlinks and macros in documents are forced to run, drawing out malicious behavior that hides until a click.
Followed URLs stay in the report as IOCs; if malicious, they are blocked, or CDR removes only the URLs and hyperlinks.

Detection case · We ran a hidden hyperlink in an APT37 ‘Artemis’ HWP document and caught concealed PowerShell activity.
This technology was selected as a 2025 KISA excellent information security technology.

Explore other technologies

Seeing it is faster than reading about it