SPECIAL NOTICE
99 -- An end-to-end adversarial input generator
- Notice Date
- 5/2/2022 9:35:10 AM
- Notice Type
- Special Notice
- Contracting Office
- ORNL UT-BATTELLE LLC-DOE CONTRACTOR Oak Ridge TN 37831 USA
- ZIP Code
- 37831
- Solicitation Number
- ORNL-TT-2022-06
- Response Due
- 5/29/2022 8:59:00 PM
- Archive Date
- 06/28/2022
- Point of Contact
- Andreana Leskovjan, Phone: 8653410433, Michael Paulus
- E-Mail Address
-
leskovjanac@ornl.gov, paulusmj@ornl.gov
(leskovjanac@ornl.gov, paulusmj@ornl.gov)
- Description
- UT-Battelle, LLC, acting under its Prime Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy for the management and operation of the Oak Ridge National Laboratory (ORNL),�is seeking a commercialization partner for new�end-to-end adversarial input generator. The ORNL Technology Transfer Office will accept license applications through May 29, 2022. The Problem:�Adversarial samples are a new class of malware specifically engineered to evade classification by machine learning�based malware detection systems. These samples constitute a serious threat because they can transfer between models, i.e., adversarial samples developed for use against a particular model architecture can achieve high success rates against completely different architectures, without needing to invest significant time and resources. Despite the threat this class of malware poses, no end-to-end systems for detecting and resolving vulnerabilities across the entire spectrum of semantic alterations exist. The Solution:�Researchers at ORNL have developed an end-to-end adversarial input generator that consumes a database of malware files and outputs a database of adversarial samples. The developed technologies have been implemented as a modular system that not only incorporates major types of semantic alterations but also allows new alteration types to be added in as they are developed. Applications:�Evaluation and enhancement of new or existing cyber security tools and products; Development of robust machine learning Advantages:�Capable of generating large quantities of evasive malware samples; Can detect malware detection system weaknesses, as opposed to generic vulnerabilities that are common across the industry; Produces a report of evasive samples that bypassed the tested tool and what alteration worked; Immediately provides tested evasive samples to vulnerability detection for retraining detection models License applications will be evaluated based on prospective partners' ability and commitment to successfully commercialize the technology, with a preference for United States-based businesses and small businesses. For a license application, contact Andreana Leskovjan, Commercialization Manager, leskovjanac@ornl.gov. ORNL Intellectual Property: ID 4695:�Functionally-Preserving Adaptive Malware Binary Rewriting Copyright ID�90000221: AMIGO
- Web Link
-
SAM.gov Permalink
(https://sam.gov/opp/43cb8b53f3634343944ee8c2d8a3042a/view)
- Record
- SN06313710-F 20220504/220502230052 (samdaily.us)
- Source
-
SAM.gov Link to This Notice
(may not be valid after Archive Date)
| FSG Index | This Issue's Index | Today's SAM Daily Index Page |