The IRMA Community
Newsletters
Research IRM
Click a keyword to search titles using our InfoSci-OnDemand powered search:
|
Architecting Trustworthy and Resilient AI Systems: Adversarial Threats, Forensic Intelligence, and Governance in Cloud-Native Environments
Abstract
Artificial intelligence (AI) and large language models (LLMs) increasingly underpin cyber defense workflows triaging alerts, correlating telemetry, classifying malware, generating incident narratives, and accelerating analyst decision-making. Yet the same properties that make AI operationally valuable (learning from data, generalizing across contexts, and automating at scale) also expand the attack surface. Adversaries can poison training data, implant hidden backdoors, steal model behavior through APIs, induce privacy leakage, or exploit prompt-manipulation weaknesses in LLM applications. When these models are deployed cloud-natively via containerized inference services, retrieval-augmented generation (RAG), agentic toolchains, and continuous delivery pipelines attack vectors multiply across datasets, MLOps supply chains, identity layers, orchestration planes, and third-party dependencies. This chapter presents a unified framework for architecting trustworthy and resilient AI systems under adversarial pressure. We synthesize adversarial AI threat models.
Related Content
|
Parth Nagar, Srinath M. S..
© 2027.
48 pages.
|
|
Swapnali Pravin Gaikwad, Saurabh Vinayak Hembade.
© 2027.
36 pages.
|
|
Titiksha Tulsidas Bhagat, Shweta Bondre, Vipin Bondre, Uma Yadav, Priya Dasarwar.
© 2027.
26 pages.
|
|
Anshik Kumar Tiwari, Brindha Subburaj.
© 2027.
22 pages.
|
|
Grace Shalini T., Pratham Shrivastav, Parthiv Gopa.
© 2027.
36 pages.
|
|
S. Aarthi, Jaypalsinh A. Gohil.
© 2027.
30 pages.
|
|
Arul Selvam P., Tamije Selvy P..
© 2027.
30 pages.
|
|
|