Sentiment Analysis of Hacker Forums with Deep Learning to Predict Potential Cyberattacks

The rapid expansion of internet communication technologies has led to the rise of popular platforms for cybersecurity discussions. Malicious hackers often brag, celebrate their hacking achievements, and issue threats on online forums. Similarly, defensive teams may post about security issues, vulnerabilities, and logs, which may contain useful information for attackers. Analyzing these communications can provide a better understanding of hacker motivations, tactics, and intentions. The information can then be used to predict future attacks and develop more effective security measures. However, manually analyzing large volumes of data is time-consuming and labor-intensive. The use of advanced techniques in neural networks has been shown to be highly effective to automatically extract information from such unstructured data. In this work, we apply a Deep Neural Network (DNN) to datasets containing hacker communications collected from different underground and surface online platforms to understand the malicious intent of hackers. We used publicly available datasets and balanced the classes by using random oversampling techniques before splitting the data into training, validation, and test sets. The data was then preprocessed with text normalization techniques, stop words removal, tokenization, stemming, and word embedding using GloVe. Our model based on the Long Short-Term Memory (LSTM) neural network architecture achieved great performance with training and validation accuracies of 99.93% and 97.48%, respectively. The ability to process and understand hacker sentiment can help organizations prepare for potential future attacks.