Adversarial Machine Learning Postdoctoral Scholar

Berkeley Lab - California - The (DST) department in the has an immediate opening for a post-doctoral researcher to perform research and development in adversarial machine learning methods for complex control systems driven by reinforcement learning. The goal of this position is to research adversarial machine learning methods that will enable safer operation of automated, adaptive deep-learning-driven cyber-physical system processes. Numerous DOE-relevant processes are becoming automated and adaptive, using machine learning techniques, such as reinforcement learning, in which the state after one run of the process automatically influences the actions taken on subsequent runs often without a human in the loop. Examples of such processes relevant to DOE today include: intelligent transportation systems, adaptive control of grid-attached equipment to stabilize power grid function, and more. This creates a vulnerability for a cyber attacker to sabotage processes through tainted training data or specially crafted inputs. To accomplish this, the postdoc will use tools from AI to develop characterizations of dynamic systems, will develop the underlying math for adversarial manipulation of the ML models, and methods that will seek to enable detection and prevention of manipulation via attacking the learning methods. In the context of the power grid, the postdoc will examine ways in which automated power grid systems can be attacked, and will develop algorithms to determine which cyber-physical system properties should be adjusted to re-stabilize unstable operating points. Important qualities of the position include experience in adversarial learning and deep learning, experience with control theory, and experience with software development, including experience in working in team-based development environments; and a strong interest in science, enabling scientific research, learning new scientific domains, and working closely with domain scientists. The department at Berkeley Lab develops software and tools to enable scientists to address complex and large-scale computing and data analysis problems beyond what is possible today. DST engages in partnerships with scientists to understand their computing and data analysis challenges to develop leading-edge solutions. Our research areas address aspects of scientific computing that are not adequately addressed by existing frameworks and tools. Details on current and recent projects are available on and . What You Will Do: Write scientific research papers suitable for submission to peer-reviewed computer science venues, such as ICML, ICLR, or NeurIPS; computer security venues such as the IEEE Symposium on Security & Privacy or the USENIX Security Symposium; or control theory venues. Use tools from AI to develop reduced order characterizations of high-dimensional dynamic systems. Develop the math underlying adversarial manipulation of the ML models and methods that will seek to enable detection and prevention of manipulation via attacking the learning methods. In the context of the power grid, examine ways in which automated power grid systems can be attacked, and will develop algorithms to determine which cyber-physical system properties should be adjusted to re-stabilize unstable operating points. Work with research staff in the Berkeley Lab Data Science & Technology Department and Grid Integration Group, with researchers and application scientists throughout the Berkeley Lab and the DOE Office of Science community, and with faculty and student collaborators from universities throughout the world. What is Required: Ph.D. degree in Computer Science, Computer Engineering, Electrical Engineering, Mathematics, Statistics, or a related technical field is required. An established track record of peer-reviewed publications in deep learning and/or adversarial machine learning. Experience with key tools used in scientific data discovery, such as Jupyter notebooks, Spark, PyML, TensorFlow, and/or related software systems. Proven experience writing software and proficiency and experience in programming languages such as C/C++ and/or Python. Demonstrated ability to work independently and collaboratively in a diverse interdisciplinary team and contribute to an active intellectual environment. Excellent written and verbal skills. Keen interest in solving science challenges. Desired Qualifications: Experience in control theory and/or signal processing. Experience with electric power systems is a plus. Proficiency with UNIX tools and computer systems. Notes: This is a full-time 2 year, postdoctoral appointment with the possibility of renewal based upon satisfactory job performance, continuing availability of funds and ongoing operational needs. You must have less than 3 years of paid postdoctoral experience. Salary for Postdoctoral positions depends on years of experience post-degree. This position is represented by a union for collective bargaining purposes. Salary will be predetermined based on postdoctoral step rates. This position may be subject to a background check. Any convictions will be evaluated to determine if they directly relate to the responsibilities and requirements of the position. Having a conviction history will not automatically disqualify an applicant from being considered for employment. Work will be primarily performed at Lawrence Berkeley National Lab, 1 Cyclotron Road, Berkeley, CA. Equal Employment Opportunity: Berkeley Lab is an Equal Opportunity/Affirmative Action Employer. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, disability, age, or protected veteran status. Berkeley Lab is in compliance with the Pay Transparency Nondiscrimination Provision under 41 CFR 60-1.4. Click to view the poster and supplement: "Equal Employment Opportunity is the Law." Lawrence Berkeley National Laboratory encourages applications from women, minorities, veterans, and other underrepresented groups presently considering scientific research careers.... - Permanent - Full-time

source http://jobviewtrack.com/en-in/job-1d1c417d42171b0a4817540c1a00024728411700001d4f0b71484c1601074913222e0902060e0b4554240c12584554434a663f015400440c0b150115044c543b0a1b45475c5f2f34001d53104f001c0e1c060900380d08014442534a2f34001d53104f001c0e1c060921475f59/69c41763f6826d73c6ab1579158a26d8.html?affid=f584d43114bf1954a48e3ec6be21b6ec