Malware Triage and Evasion Analysis Workbench

A malware-analysis workbench for classifying suspicious samples, identifying evasion logic, and documenting behavior under controlled execution.

Impact

Classified suspicious samples through static and dynamic indicators.

Impact

Documented anti-analysis behavior and execution preconditions.

Impact

Separated benign automation artifacts from higher-risk malware behaviors.

Deliverables

  • Triage checklist
  • Indicator extraction scripts
  • Evasion and behavior report

Technical Overview

Suspicious binaries and scripts often contain partial indicators, environment checks, staged payload logic, and noisy false leads. The goal was to build a disciplined triage workflow that surfaces useful evidence quickly.

Architecture

flowchart TD sample_file[Suspicious sample] --> static_scan[Static features] sample_file --> controlled_run[Controlled execution] static_scan --> indicators[Indicators] controlled_run --> observed_behavior[Observed behavior] observed_behavior --> risk_score[Capability score] indicators --> risk_score risk_score --> triage_report[Triage decision]

Model

T=0.40B+0.25E+0.20N+0.15PT = 0.40B + 0.25E + 0.20N + 0.15P

where T is triage priority, B is observed behavior, E is evasion strength, N is network activity, and P is persistence capability.

Implementation Sketch

def score_sample(features):
    weights = {'network': 0.20, 'persistence': 0.15, 'evasion': 0.25, 'behavior': 0.40}
    return sum(features.get(k, 0) * w for k, w in weights.items())

if sample.has_vm_check or sample.sleeps_long:
    notes.append('anti-analysis gate requires alternate run profile')

Engineering Approach

  • Start with hash, metadata, strings, imports, and packer indicators.
  • Run controlled dynamic tests only after defining expected observations and rollback boundaries.
  • Score behavior by capability, reachability, and confidence rather than by isolated suspicious strings.

Results

  • Produced concise behavior summaries for suspicious files, web payloads, and scripted malware.
  • Identified evasion gates such as VM checks, sleep delays, locale checks, and network reachability tests.
  • Made the analysis reproducible through command logs, extracted indicators, and scoped recommendations.

What This Demonstrates

  • A reliable triage process values reproducibility over speculation.
  • Evasion analysis should document both what executed and what conditions prevented execution.
  • Indicator severity is strongest when static and dynamic evidence agree.