一键导入
run-attacks
Run the pac_clip_sum membership inference attack test suite and summarize results.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Run the pac_clip_sum membership inference attack test suite and summarize results.
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
Reference material for differential privacy concepts. Auto-loaded when discussing privacy, attacks, sensitivity, clipping, local/smooth sensitivity, elastic sensitivity, instance-based DP, per-instance DP, individual privacy accounting, personalized DP, Pufferfish/Blowfish, or DP relaxations.
Reference material for PAC privacy internals. Auto-loaded when discussing PAC mechanism, noise, counters, or clipping.
Reference for PAC DDL syntax — PRIVACY_KEY, PRIVACY_LINK, PROTECTED, SET PU, and the parser. Auto-loaded when discussing table setup, privacy units, or protected columns.
Build and run pac_clip_sum unit tests.
基于 SOC 职业分类
| name | run-attacks |
| description | Run the pac_clip_sum membership inference attack test suite and summarize results. |
PAC (Probably Approximately Correct) privacy privatizes SQL aggregates via 64 parallel SWAR bitslice counters with MI-bounded noise. pac_clip_sum adds per-user contribution clipping using magnitude-level decomposition (4x bands, 2-bit shift) with distinct-contributor bitmaps. Unsupported outlier levels are hard-zeroed to prevent variance side-channel attacks.
GEN=ninja make 2>&1 | tail -5bash attacks/clip_attack_test.sh 2>/dev/nullbash attacks/clip_multirow_test.sh 2>/dev/nullbash attacks/clip_hardzero_stress.sh 2>/dev/nullSummarize results as a table:
attacks/clip_attack_results.md