AI Supply Chain Attack: Breaching Through the Backdoor
Course
Shoumik Chandra of Secvera Technologies demonstrates real-world LLM supply chain attacks - from tokenizer backdoors that evaded Hugging Face security scans to RAG-based exploits on public chatbots - and shares actionable detection strategies.
As organizations race to embed large language models into chatbots, internal agents and critical workflows, a new class of supply chain attack has emerged that traditional software security controls were never designed to stop. The artificial intelligence supply chain - spanning pre-training data, base model weights, ML frameworks, fine-tuning adapters, RAG vector stores and MCP server tools - presents a distinct attack surface at every layer.
In this session, led by Shoumik Chandra, cybersecurity manager at Secvera Technologies, you will learn:
- How attackers exploit every layer of the LLM supply chain;
- How a RAG-based attack on a public chatbot achieved system access with zero prior knowledge;
- Practical, actionable detection strategies to validate AI models before production deployment.
Here is the course outline:
AI Supply Chain Attack: Breaching Through the Backdoor |
Completion
The following certificates are awarded when the course is completed:
 |
CPE Credit Certificate |
