SAMPA: Security Analysis and Monitoring to Prevent Abuse of High Performance Computing Environments
- Intelligent Automation Inc. (IAI)
- University of South Florida (USF)
HPC systems should not be abused to carry out cybercrimes or execute improper jobs. Existing defense solutions such as fingerprinting communication and computation require runtime data collection and processing, which inevitably impacts the performance of HPC. Moreover, existing solutions are ineffective against adversarial scenarios where malicious users intend to hide their behavior patterns. Thus, it is essential to build a powerful offline software analysis tool that can scan/analyze obfuscated HPC programs to detect illicit behaviors. In Phase I, we designed and developed a prototype of deep-learning-based analysis toolkit dubbed SAMPA. It classifies user-submitted program binaries into their classes and provides a similarity score between a given program and any programs that are not allowed to run on HPC. With SAMPA, the abuse of HPC systems can be prevented without introducing runtime overhead to HPC systems.
- Research Organization:
- Intelligent Automation Inc.
- Sponsoring Organization:
- USDOE Office of Science (SC)
- Contributing Organization:
- University of South Florida (USF)
- DOE Contract Number:
- SC0018476
- OSTI ID:
- 1492551
- Type / Phase:
- STTR (Phase I)
- Previous Contract Number:
- DE-SC0018476
- Report Number(s):
- FINAL_IAI_2356_18476
- Country of Publication:
- United States
- Language:
- English
Similar Records
Spy the Lie: Detecting Malicious Insiders
Final Report- HOST: HPC Obfuscation and Security Toolkit