Diagnosing Vision-and-Language Navigation: What Really Matters

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 587 - 599 July 10-15, 2022 ©2022 Association for Computational Linguistics A Word is Worth A Thousand Dollars: Adversarial Attack on Tweets Fools Stock Prediction Yong Xie1, Dakuo Wang2,†, Pin-Yu Chen2, Jinjun Xiong3, Sijia Liu2,4 and Sanmi Koyejo1 1University of Illinois Urbana-Champaign, 2IBM 3State University of New York at Buffalo, 4Michigan State University Abstract More and more investors and machine learn- ing models rely on social media (e.g., Twit- ter and Reddit) to gather real-time informa- tion and sentiment to predict stock price move- ments. Although text-based models are known to be vulnerable to adversarial attacks, whether stock prediction models have similar vulnera- bility is underexplored. In this paper, we exper- iment with a variety of adversarial attack con- figurations to fool three stock prediction vic- tim models. We address the task of adversarial generation by solving combinatorial optimiza- tion problems with semantics and budget con- straints. Our results show that the proposed attack method can achieve consistent success rates and cause significant monetary loss in trading simulation by simply concatenating a perturbed but semantically similar tweet. 1 Introduction The advance of deep learning based language mod- els are playing a more and more important role in the financial context, including convolutional neutral network (CNN) (Ding et al., 2015), recur- rent neutral network (RNN) (Minh et al., 2018), long short-term memory network (LSTM) (Hiew et al., 2019; Sawhney et al., 2021; Hochreiter and Schmidhuber, 1997), graph neutral network (GNN) (Sawhney et al., 2020a,b), transformer (Yang et al., 2020), autoencoder (Xu and Cohen, 2018), etc. For example, Antweiler and Frank (2004) find that com- ments on Yahoo Finance can predict stock market volatility after controlling the effect of news. Cook- son and Niessner (2020) also show that sentiment disagreement on Stocktwits is highly related to cer- tain market activities. Readers can refer to these survey papers for more details (Dang et al., 2020; Zhang et al., 2018; Xing et al., 2018). † Corresponding author dakuo@acm.org. Our code is available at https://github.com/yonxie/ AdvFinTweet Figure 1: An example of word-replacement adversarial attack. (Top) benign tweet leads Stocknet to pre- dict stock going UP; (Bottom) adding an adversarial retweet leads Stocknet to predict stock going DOWN. It is now known that text-based deep learning models can be vulnerable to adversarial attacks (Szegedy et al., 2014; Goodfellow et al., 2015). The perturbation can be at the sentence level (e.g., Xu et al., 2021; Iyyer et al., 2018; Ribeiro et al., 2018), the word level (e.g., Zhang et al., 2019; Alzantot et al., 2018; Zang et al., 2020; Jin et al., 2020; Lei et al., 2019; Zhang et al., 2021; Lin et al., 2021), or both (Chen et al., 2021). We are interested in whether such adversarial attack vulnerability also exists in stock prediction models, as these models embrace more and more human-generated media data (e.g., Twitter, Reddit, Stocktwit, Yahoo News (Xu and Cohen, 2018; Sawhney …

Content truncated to 500 words