Dataset: Serverless Forensics — Welch's t-test Compromise Detection

Serverless Forensics: Welch's t-test Dataset Dataset supporting the paper "Serverless Forensics: Automated Compromise Detection via Welch's t-test" (Mansour, Shanmugam, Yeo). FILEserverless_forensics_final.csv — 96 rows, one per Lambda invocation. COLUMNS- group: Experimental condition, either "compromised" or "non-compromised"- attack_scenario: One of four OWASP-derived attack scenarios — Event Injection, Broken Authentication, Sensitive Data Exposure, Security Misconfiguration- duration_ms: AWS Lambda execution duration (ms), captured via CloudWatch — the primary dependent variable analyzed with Welch's t-test- avg_incidence_rate: Average CWE incidence rate (%) for the corresponding attack category, from the MITRE CWE / NVD mapping (see Table 1 of the paper). Constant across all 12 rows within a scenario — this is a scenario-level value, not per-invocation.- total_occurrences: Total NVD vulnerability record count for the corresponding CWE cluster, as of the query date (see Table 1 of the paper). Also scenario-level, not per-invocation. STRUCTUREn = 12 independent invocations per condition (compromised / non-compromised) per attack scenario, across 4 scenarios = 96 total observations. REPRODUCING THE ANALYSISFor each attack scenario, a Welch's t-test (unequal-variance, two-tailed) was computed comparing duration_ms between the compromised and non-compromised groups, with Bonferroni correction (alpha = 0.05/4 = 0.0125) applied across the four scenarios. See Sections 3.4 and 4.6-4.7 of the paper for full methodology. LICENSECC BY 4.0