Ai security papers guide

Skill brycewang-stanford/Auto-Empirical-Research-Skills/skills/43-wentorai-research-plugins/skills/domains/cs/ai-security-papers-guide

AI security papers from top-4 security conferencesFrom its SKILL.md

Install

npx -y skills add brycewang-stanford/Auto-Empirical-Research-Skills --skill ai-security-papers-guide

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

no licenseNo license file was found in the repository. Code published without one is not open source by default, so using it at work is a question for whoever answers licensing questions where you are.

SKILL.md

3.6 KB, 837 tokens by cl100k_base, as published. Nobody here has run it

AI Security Papers Guide (BIG4 Venues)

Overview

A curated collection of AI security papers from the top-4 security conferences: IEEE S&P, ACM CCS, USENIX Security, and NDSS. Covers adversarial attacks, model stealing, data poisoning, privacy attacks, deepfake detection, and LLM security. Organized by year and venue, focusing exclusively on peer-reviewed work from these prestigious venues.

Venues

Venue	Full Name	Focus
S&P	IEEE Symposium on Security and Privacy	Broad security + privacy
CCS	ACM Conference on Computer and Communications Security	Systems security
USENIX	USENIX Security Symposium	Systems + network security
NDSS	Network and Distributed System Security	Network security

Topic Categories

AI Security (BIG4)
├── Adversarial ML
│   ├── Evasion attacks (adversarial examples)
│   ├── Poisoning attacks (backdoors, trojans)
│   ├── Model stealing (extraction, distillation)
│   └── Defenses (certified robustness, detection)
├── Privacy Attacks
│   ├── Membership inference
│   ├── Model inversion
│   ├── Attribute inference
│   └── Training data extraction
├── LLM Security
│   ├── Prompt injection
│   ├── Jailbreaking
│   ├── Data leakage
│   └── Alignment attacks
├── Deepfakes
│   ├── Generation methods
│   ├── Detection techniques
│   └── Watermarking
└── Federated Learning Security
    ├── Byzantine attacks
    ├── Gradient leakage
    └── Secure aggregation

Key Papers by Year

# Recent highlights
papers_2024_2025 = [
    {"title": "Not What You've Signed Up For: "
              "Compromising Real-World LLM-Integrated Applications",
     "venue": "S&P 2024", "topic": "LLM security"},
    {"title": "Prompt Stealing Attacks Against "
              "Text-to-Image Generation Models",
     "venue": "S&P 2024", "topic": "Prompt extraction"},
    {"title": "Backdoor Attacks on Language Models",
     "venue": "CCS 2024", "topic": "NLP backdoors"},
    {"title": "Membership Inference in LLMs",
     "venue": "USENIX 2024", "topic": "Privacy"},
]

for p in papers_2024_2025:
    print(f"[{p['venue']}] {p['title']}")
    print(f"  Topic: {p['topic']}")

Research Trends

### Emerging Areas (2024-2025)
1. **LLM security** — Jailbreaking, prompt injection, agent attacks
2. **Supply chain attacks** — Poisoned models, malicious packages
3. **Multi-modal attacks** — Cross-modal adversarial examples
4. **Agent security** — Attacks on LLM-based autonomous systems
5. **Watermarking** — LLM output detection, IP protection
6. **Unlearning** — Machine unlearning verification and attacks

Use Cases

Security research: Find state-of-the-art attack/defense methods
Threat modeling: Understand AI system vulnerabilities
Literature review: Systematic coverage of BIG4 AI security
Course material: Graduate-level AI security curriculum
Red teaming: Learn evaluation techniques for AI systems

References

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most security skills give in 837 tokens

Counted across 666 of the 889 authors here whose files we hold, read 2026-09-06

Use parameterized queries for database accessin 82 of 666, across 79 files
Hash passwords with BCryptin 55 of 666, across 39 files
Implement rate limiting for public endpointsin 48 of 666, across 34 files
Use environment variables for secretsin 35 of 666
Scan dependencies for vulnerabilitiesin 35 of 666, across 24 files
Validate and sanitize all user inputin 35 of 666, across 32 files
Add security headers to all responsesin 34 of 666, across 20 files
Validate all external input at the system boundaryin 26 of 666, across 25 files
Use parameterized queries to prevent SQL injectionin 25 of 666, across 13 files
Store secrets in Vault or environment variablesin 25 of 666, across 10 files
Run containers as a non-root userin 21 of 666, across 18 files
Validate all input using Bean Validationin 19 of 666, across 5 files

Said here and by no other author read

Focus exclusively on peer-reviewed papers from BIG4 venues
Categorize research by adversarial ML and privacy attacks
Use BIG4 papers for threat modeling AI systems
Apply research findings to red teaming AI models

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.