Early Detection of DDoS Attacks Using Photonic Neural Networks

Deep Learning (DL) has been extensively used in challenging tasks including security applications such as Distributed Denial of Service (DDoS) attacks. However, the high speed requirements of such applications along with the high complexity of DL models restrict the practical use of DL in real systems. Photonic neuromorphic hardware provides several advantages over electronic counterparts since it can operate at very high frequencies with lower power consumption. To this end, in this paper, we propose employing a photonic neuromorphic lookaside accelerator, aiming to perform real-time traffic inspection, enabling us to detect port-scanning attacks, which are indicative of DDoS attacks. We have designed, trained, and evaluated a Photonic Neural Network (PNN) capable of detecting DDoS attacks and operating on such photonic neuromorphic lookaside accelerators. The experimental evaluation is performed on Transport Control Protocol (TCP) traces obtained by simulating a port scanning attack and demonstrates the effectiveness of the proposed approach.