2nd KDD Workshop on Anomaly Detection in Finance



Anchorage, Alaska, USA

August 5, 2019

Summary

Detecting anomalies and novel events is vital to the financial industry. These events are often indicative of illegal activities such as credit card fraud, identity theft, network intrusion and money laundering. Left unchecked, these activities can cause poor customer experiences and billions of dollars in losses. In addition to these activities, a new threat is emerging in the form of fake news in financial media outlets that can lead to distortions in trading strategies and investment decisions. A number of new techniques are emerging to tackle these problems including semi-supervised learning methods, deep learning algorithms, network/graph based solutions as well as linguistic approaches. These methods must often be able to work in real-time and be able to handle large volumes of data. The purpose of this workshop is to bring together researchers and practitioners to discuss both the problems faced by the financial industry and potential solutions.

Call for Papers

We invite papers on anomaly and novelty detection with applications for the financial industry. Topics of interest include, but are not limited to, the following:

Business problems:

● Financial Crimes:

      • Anti-money laundering
      • Fraudulent transactions
      • Identity theft and fake account registration
      • Promotion credit abuse
      • Account takeover
      • Insurance fraud

● Risk Modeling

      • Enhanced risk modeling
      • Regulation-aware feature engineering
      • Monitoring

● Other applications

      • Fake news
      • Social media mining
      • Early detection of emerging phenomena
      • Manipulation in cryptocurrency markets

Technical problems:

● Semi-supervised anomaly detection (aka Novelty Detection):

      • Data available for training does not contain any anomalies and represents expected operation of the system.
      • The algorithm classifies everything that does not fit in description of the previously seen data as a “novelty”.

● Unsupervised anomaly detection (aka Outlier Detection):

      • Data available for training may contain anomalies, which are assumed to be rare.
      • Anomalies detected by the algorithm are considered to be “outliers” relative to the majority of available data.

● Explainable models for anomaly detection

      • Models that can explain their decisions in interpretable ways
      • Post-hoc methods that can be used to explain outputs of other detection algorithms

● Human-in-the-loop anomaly detection

      • Interactive ranking techniques
      • Methods that can handle exploration-exploitation trade-off
      • Novel human feedback gathering strategies beyond labeling

● Adversarially-robust detection

      • Methods that are provably robust to evasion and camouflage
      • Evasion-cost aware fraud and intrusion detection
      • Analysis of evasion schemes and camouflage mechanisms


We also invite tutorials and introductory papers to bridge the gap between academia and the financial industry:

Overview of Industry Challenges

  • Short papers from financial industry practitioners that introduce domain specific problems and challenges to academic researchers. These papers should describe problems that can inspire new research directions in academia, and should serve to bridge the information gap between academia and the financial industry.

Algorithmic Tutorials

  • Short tutorials from academic researchers that explain current solutions to challenges related to anomaly detection, not necessarily limited to the financial domain. These tutorials will serve as an introduction and enable financial industry practitioners to employ/adapt latest academic research to their use-cases.


Submission Guidelines:

All submissions must be PDFs formatted in the Standard ACM Conference Proceedings Template. Submissions are limited to 8 content pages or less, including all figures and tables but excluding references. Due to popular request we are also accepting 1-page summaries or extended abstracts. All accepted papers will be presented as posters; some may be selected for oral presentations, depending on schedule constraints. Accepted papers will be posted on the workshop website or, at the authors’ request, may be linked to an external repository such as arXiv.

Authors should clearly indicate in their abstracts the kinds of submissions that the papers belong to, to help reviewers better understand their contributions.


Papers (full papers or 1-page summaries) should be submitted on CMT3 by May 28, 2019 11:59 PM Pacific Time

https://cmt3.research.microsoft.com/ADF2019


Key dates