July 2022
ICML-ML4Cyber
ICML Workshop on Machine Learning for Cybersecurity
The primary aim of this workshop is to build a mutual comprehensive awareness of the problem and solution spaces across the greater ML community and the Cybersecurity/ML for Cybersecurity communities.
Call for Papers
Submission Deadline 29-MAY-2022 AOE
Submission Deadline 29-MAY-2022 AOE
Overview
Following a series of crippling cyber-attacks that targeted a majority of public and social sectors – including schools, hospitals, critical infrastructure, and private businesses – the global community has increased its attention on the wider societal impacts of major cybersecurity events, forming task forces like the UN Open Ended Working Group on Cyber Security and undertaking policy efforts to mitigate these impacts. Such actions are important, but policy changes only represent one side of the solution. On the other are technical developments, within which machine learning (ML) has been proposed as a key component of future cyber defense tools, requiring rapid development to provide the speed and scale needed to detect and respond to new and emerging cybersecurity threats.
Cybersecurity itself is inherently a systems problem and piecewise application of off-the-shelf ML tools leaves critical gaps in both sophistication and interpretable context needed for comprehensive security systems. To successfully develop ML-based cybersecurity defenses, a greater degree of cross-pollination across the ML and cybersecurity communities is needed because both are highly specialized technical domains. Moreover, the requisite ML topics needed to successfully leverage ML for cybersecurity – such as time series analytics, game theory, deep learning, reinforcement learning, representation learning, semi-supervised and self-supervised learning, learning on large scale streaming data, interpretable and robust autonomous systems, etc. – are foundational to the ML research community community.
The ICML workshop on Machine Learning for Cybersecurity (ML4Cyber) strives to push the state-of-the-art in ML-based cybersecurity defenses by increasing this cross-pollination, building a mutual comprehensive awareness of the problem and solution spaces across the greater ML research and cybersecurity/ML-for-cybersecurity communities. In its inaugural year, the ML4Cyber workshop will provide meaningful engagement to help bring these communities together, including a general session intended for ML researchers covering the core tenets of cybersecurity and available tools and datasets, a contributed session of technical papers, and panel discussions discussing hard problems and new developments in cybersecurity, ML, and their intersection.
Topics of Interest
We broadly seek contributions looking at ML as applied to cybersecurity, with preference given to those papers that discuss the following:
Hard ML problems applied to cybersecurity, including but not limited to:
Extremely fast content drift (minutes to hours);
Extreme class imbalance (on the order of 1:100000 malicious to benign); and
Continual learning across multiple timescales.
Advances to state of the art AI/ML conceived through novel application to cybersecurity
Autonomous or semi-autonomous cyber agents that can operate at the speed, scale, stealth, and ingenuity of cyber threat actors.
Robust model training, deployment, and performance evaluation as applied to the cybersecurity domain, which is inherently adversarial.
Cybersecurity specific use cases of ML
Use case scenarios with new/novel public data set release.
High priority gaps in cyber defense and case studies of how ML solutions are being applied.
Trustworthy AI/ML for Cybersecurity
Applications of explainability, causality, robustness, etc. which aid in building trust with human cyber defenders.
Privacy preserving and/or federated ML that can enable public data sharing and/or cross-organizational cyber response -- such as training ML methods for distributed threat hunting, sharing of cyber threat intelligence data, botnet takedowns, etc. -- while reducing risk and maintaining policy compliance.
ML driven by cyber situational awareness that can provide meaningful output beyond normal/anomalous.
Submission Guidelines
Authors may submit one of the following types of papers:
Research papers (up to 8 pages, +1 for references) in the area of applying ML to cybersecurity.
Position papers (up to 4 pages, +1 for references) describing early results or hard problems.
Use cases (up to 4 pages) describing experiences fielding ML approaches within cybersecurity.
Survey papers (up to 8 pages, + 1 for references) providing a comprehensive overview of previously published research on a problem within the intersection of ML and cybersecurity.
The submission deadline is 22-MAY-2022 AOE. Submit papers using this site. While the ML4Cyber workshop solicits all four types of contributions, preference will be given to original research papers covering a topic described in the list above. All papers should follow ICML formatting guidelines and be submitted in PDF form, anonymized, and in English.
Workshop Organizers
Dr. Andy Applebaum
Apple
Prof. William Arbaugh
Five Directions
Jack W. Davidson
University of Virginia
Joseph Edappully
NSA LACR
Dr. John Emanuello
NSA LACR
Dr. Howie Huang
George Washington University
Andrew Golczynski
NSA LACR
Dr. Nicole Nichols
Apple
Mr. Tejas Patel
DARPA
Dr. Ahmad Ridley
NSA
Dr. Vance Wong
NSA LACR
History: Despite ample participation from the ML for cybersecurity community at NeurIPS and ICML, a workshop on this emerging domain has not been hosted at either of these premier ML conferences. Venues which have provided some forum for discussions, with varying emphasis on human or ML based defenses for cyber security, include SIAM Data Mining Conference 2021, AAAI-AICS, CAMLIS, USENIX, IJCAI-ACD, IEEE-Security and Privacy, ACM-CCS, etc