ML Security Papers
Browse Stats

Research landscape

4030 papers across 20 OWASP categories

4030
papers
1255
attack (31%)
1905
defense (47%)
389
this month

Monthly publications

OWASP ML Top 10

ML01 Input Manipulation Attack
920
ML02 Data Poisoning Attack
315
ML03 Model Inversion Attack
269
ML04 Membership Inference Attack
226
ML05 Model Theft
175
ML06 AI Supply Chain Attacks
62
ML07 Transfer Learning Attack
94
ML08 Model Skewing
9
ML09 Output Integrity Attack
825
ML10 Model Poisoning
369

OWASP LLM Top 10

LLM01 Prompt Injection
1367
LLM02 Insecure Output Handling
0
LLM03 Training Data Poisoning
51
LLM04 Model Denial of Service
23
LLM05 Supply Chain Vulnerabilities
0
LLM06 Sensitive Information Disclosure
233
LLM07 Insecure Plugin Design
137
LLM08 Excessive Agency
258
LLM09 Overreliance
0
LLM10 Model Theft
58

LLMs for Security

Research on using LLMs as offensive or defensive security tools (orthogonal to OWASP).

LS01 Vulnerability Discovery
10
LS02 Exploit Generation
5
LS03 Reconnaissance & OSINT
1
LS04 Patch & Remediation
2
LS05 Triage & Prioritization
2
LS06 Red-Team Agents
63
LS07 Blue-Team Agents
22
LS08 Reverse Engineering
1
LS09 Fuzzing & Test Generation
2
LS10 Benchmarks & Evaluation
53

Domains

nlp 2239
vision 1634
multimodal 640
generative 480
federated-learning 285
audio 194
tabular 102
graph 95
reinforcement-learning 80
timeseries 41

Paper types