Bezpieczeństwo sztucznej inteligencji
- Obaj prelegenci powinni przetestować konfigurację rzutnika przed rozpoczęciem zajęć, aby zmiana prelegenta odbyła się sprawnie.
- Każda prezentacja powinna trwać 35 minut.
- Slajdy prezentacji powinny być przygotowane w języku angielskim, natomiast językiem prezentacji jest język polski.
- Slajdy prezentacji powinny być wysłane do prowadzącego na 24h przed rozpoczęciem zajęć.
Slajdy z pierwszych wprowadzających zajęć
Propozycje tematów
General
Introduction into AI Security- Tools review and main libraries
Data sets review- Building playground and show case for (maching learning | deep learning |
LLMs) Agentic AI security
Offensive security
AI in misinformationAI in socialengeneeringAI in creating malwareAI in phishing e-mailsAI in deep fakes
Defensive capabilities
AI in penetration testing overviewAI in cyber threat analysisAI in malware detectionAI in software vulnerabilities detection & repairAI in secure code generation- AI in test cases generation
AI in phishing detectionAI in code obsfuscation and deobfuscation- AI in web content filtering
- AI in software bug-fixing
AI in data classificationAI in incident detection & responseAI in watermarking
Vulnerabilities in AI solutions/models
OWASP Top 10 for LLMsHalucinationsData poisoningPrompt injectionsPrivacy challengesAdversarial attacksModel inversion attaks
Harmonogram
| Data | Slot | Temat | Prelegent | Materiały |
|---|---|---|---|---|
| Seria 1 | ||||
| 15.10 | 1 i 2 | Introduction into AI Security | Szymon L. | ... |
| 22.10 | 1 | OWASP Top 10 for LLMs | Mateusz K-s. | ... |
| 2 | Halucinations | Julia K. | ... | |
| 29.10 | 1 | AI in creating malware | Miriam B. | ... |
| 2 | Model inversion attaks | Mateusz K-t. | ... | |
| 3 | Data sets review | Kacper G. | ... | |
| 05.11 | 1 | AI in misinformation | Michał W. | ... |
| 2 | Agentic AI security | Wiktor O. | ... | |
| 12.11 | 1 | AI in deep fakes | Kornel O. | ... |
| 2 | Data poisoning | Oskar K. | ... | |
| 19.11 | - | - | - | ... |
| 03.12 | 1 | AI in phishing e-mails | Paweł N. | ... |
| 2 | Adversarial attacks | Krzysztof J. | ... | |
| 3 | Building playground and show a case for LLM | Krzysztof Ch. | ... | |
| 10.12 | 1 | Prompt injections | Adrian W. | ... |
| 2 | AI in socialengeneering | Katarzyna T. | ... | |
| Seria 2 | ||||
| 17.12 | 1 | LLMs & VLLMs Safety Pipelines | Julia K. | ... |
| 2 | AI in data classification | Oskar K. | ... | |
| 07.01 | 1 | AI in phishing detection | Kornel O. | ... |
| 2 | AI in penetration testing overview | Mateusz K-s | ... | |
| 3 | AI in secure code generation | Mateusz K-t | ... | |
| 14.01 | 1 | AI in Watermarking | Michał W. | ... |
| 2 | AI in incident detection & response | Adrian W. | ... | |
| 3 | Privacy challenges | Katarzyna T. | ... | |
| 21.01 | 1 | AI in software vulnerabilities detection & repair | Miriam B. | ... |
| 2 | AI in code obsfuscation and deobfuscation | Krzysztof Ch. | ... | |
| 28.01 | 1 | AI in malware detection | Wiktor O. | ... |
| 2 | AI in cyber threat analysis | Kacper G. | ... | |
| 04.02 | 1 | ... | ||
| 2 | ... | |||