2020/10/19 by Elior Nehemya, Yael Mathov, Nehemya, Elior +5 · 2 citations
Computer Science · #Adversarial Robustness in Machine Learning #Cryptography and Security (cs.CR) #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #FOS: Economics and business #Machine Learning (cs.LG) #Trading and Market Microstructure (q-fin.TR)
paper · pdf · doi:10.48550/arxiv.2010.09246
openalex publication_date 2020/10/19 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
In recent years, machine learning has become prevalent in numerous tasks,\nincluding algorithmic trading. Stock market traders utilize machine learning\nmodels to predict the market's behavior and execute an investment strategy\naccordingly. However, machine learning models have been shown to be susceptible\nto input manipulations called adversarial examples. Despite this risk, the\ntrading domain remains largely unexplored in the context of adversarial\nlearning. In this study, we present a realistic scenario in which an attacker\ninfluences algorithmic trading systems by using adversarial learning techniques\nto manipulate the input data stream in real time. The attacker creates a\nuniversal perturbation that is agnostic to the target model and time of use,\nwhich, when added to the input stream, remains imperceptible. We evaluate our\nattack on a real-world market data stream and target three different trading\nalgorithms. We show that when added to the input stream, our perturbation can\nfool the trading algorithms at future unseen data points, in both white-box and\nblack-box settings. Finally, we present various mitigation methods and discuss\ntheir limitations, which stem from the algorithmic trading domain. We believe\nthat these findings should serve as an alert to the finance community about the\nthreats in this area and promote further research on the risks associated with\nusing automated learning models in the trading domain.\n