Wireless Sensor Networks (WSN) are widely used across various fields. WSN is composed of many low-cost, high-performance, plug-and-play sensor nodes. WSN is used across a wide range of applications. Distributed Denial-of-Service (DDoS) attacks that overwhelm targeted resources, denying access to legitimate users. It prevents web servers from serving resources to clients. One type of DDoS is an HTTP flood attack, in which an attacker targets network resources, such as bandwidth (the amount of data a network can carry) and CPU processing (a central processing unit's ability to compute). The attacker sends multiple HTTP POST requests to the server to transmit data. In addition, the attacker sends multiple HTTP GET requests to retrieve data from the server. Previous work identified HTTP TRACE flood attacks that misuse the HTTP TRACE method. The HTTP TRACE method returns the received HTTP request, thereby exposing sensitive data. These attacks employ static URLs, which degrade the overall performance of the WSN. To address these issues, introduce the proposed method, the Enhanced Deep Spectral Multi-Layer Convolutional Neural Network (EDSMCNN), a deep learning model designed to improve CPU performance, handle multiple URL requests, and predict TRACE attack traffic based on the maximum-weighted features. First, input the HTTP flood attack dataset, which is available online. The initial step is preprocessing: analyzing and preparing data. The datasets are preprocessed to reduce the dimensionality (number of input features) of non-redundant data. Average weightage scaling feature to filter using the spider algorithm (an optimization technique based on social spider behavior) selects relevant features based on Lattice Service Rate Access Values (LSRAV, a metric evaluating service rate in the system) and observes Trace flood Traffic, considering parameters such as URL, protocol, and IP address (unique network identifier). Social spiders compare feature selection patterns with rank results. Next, SoftMax (a mathematical function that converts numbers to probabilities) generates logistic neurons using the Logistic Activation Function (SLAF, an activation mechanism for neural networks) for HTTP POST and GET requests to prevent trace attacks. Compared to standard convolutional methods, the system achieves high efficiency in detecting HTTP-TRACE flooding attacks. Experimental results show the proposed system improves CPU performance and reduces computation time, helping avoid traffic in one or more HTTP requests. In WSN, security is of paramount importance, and the emergence of novel attack vectors poses significant challenges. This abstract highlight a concerning scenario in which sensor nodes are targeted by a TRACE attacker via the injection of a backdoor entry file. Once compromised, the attacker gains unauthorized access to the WSN web server, potentially exposing sensitive data or gaining control over the network. This sophisticated attack underscores the need for robust security measures in WSNs, including intrusion detection systems, encryption, and authentication protocols, to safeguard against such threats and ensure the integrity and confidentiality of data transmitted and collected within these networks. Developing effective countermeasures to address these emerging attack vectors is crucial for the continued deployment and reliability of WSN in various critical applications.
Tamilselvi et al. (Mon,) studied this question.