Jellyfish optimized deep learning framework for cache pollution attack detection in NDN environment

Abstract

Named data networking (NDN) is a promising paradigm that replaces the traditional connection-based model with a content-based approach for future Internet infrastructures, allowing data retrieval by unique names. However, NDN faces threats like cache pollution attacks (CPA) which can lead to increased cache misses and data retrieval delays, and pose significant risks to its efficiency and security. In this paper, a novel jellyfish optimized deep learning (DL) framework for cache pollution attack detection in NDN environment (DSODAL) technique has been proposed to detect the CPA attack with high accuracy. To detect CPA in NDN, a dual-gate attention-based long short-term memory (LSTM) (DA-LSTM) network is used which is optimized using the jellyfish search optimization (JSO) algorithm. The DA-LSTM analyzes request sequences to identify malicious patterns, enhancing cache pollution detection. Nodes manage these requests using the content store (CS) for caching frequently accessed data, optimizing retrieval efficiency, and the pending interest table (PIT) to track and process incoming requests. The DA-LSTM analyzes request sequences to identify malicious patterns and detect CPA attacks. The DSODAL approach performance is evaluated using accuracy, precision, recall, F1-score, average delay time, and mean square error (MSE). The DSODAL model advances the overall accuracy by 1.74%, 2.34%, and 2.7%, over existing HCDLP, ACISE, and AHISM techniques.