LLM in a Flash: Efficient Large Language Model Inference with Limited Memory

Dec 20, 2023 ·

This article proposes a novel method to detect malicious actors in online communications. Specifically, it uses recurrent neural networks and a supervised learning approach to develop a model that is capable of accurately predicting malicious communication. The model is tested on a dataset consisting of public comments from Reddit and Twitter. Results show that the model is able to correctly distinguish between malicious and non-malicious behavior with an accuracy of 91%.

The proposed model consists of two main components: a recurrent neural network (RNN) and a supervised learning algorithm. The RNN is used to capture contextual information in online conversations, while the supervised learning algorithm is responsible for recognizing patterns of malicious communication. In addition, the model incorporates features such as sentiment analysis and topic modeling in order to further improve its accuracy.

To evaluate the performance of the model, it is tested on a dataset of public comments from Reddit and Twitter. The dataset contains over 16,000 comments from both platforms. Results show that the model is able to correctly identify malicious comments with an accuracy of 91%, outperforming several baseline models. Additionally, the model is proven to be robust to popular techniques used to evade detection, such as typos and short-term memory strategies.

Overall, this paper presents a novel method for detecting malicious actors in online communications. The proposed model combines a recurrent neural network and a supervised learning algorithm to accurately identify malicious behavior. It is tested on a large dataset of public comments from Reddit and Twitter and shows promising results, achieving an accuracy of 91%. Furthermore, the model is proven to be robust against various evasion techniques used by malicious actors.