Hybrid Deep Learning Model for Multiple Cache Side Channel Attacks Detection: A Comparative Analysis

Cache side-channel attacks have emerged as a sophisticated and persistent threat, capable of extracting sensitive user information by exploiting vulnerabilities in modern processors. These attacks leverage the inherent weaknesses in shared computational resources, especially the last-level cache. They infer patterns in data access and execution flows, often bypassing traditional security defenses. These are particularly dangerous because they require no physical access to the victim's device, making remote attacks feasible. This study focuses on a specific class of these threats- fingerprinting attacks -where an adversary can use cache-side channels to monitor and analyze the behavior of co-located processes, potentially revealing confidential information such as encryption keys or user activity patterns. A comprehensive threat model illustrates how an attacker, sharing computational resources with a target system, can exploit these side-channels. This exploitation allows the attacker to learn patterns in data access, potentially compromising sensitive information. To mitigate such risks, a hybrid deep learning model is proposed to detect cache side-channel attacks. Its performance is compared with five widely-used Deep Learning models like Multi Layer Perceptron, Convolutional Neural Network, Simple Recurrent Neural Network, Long Short Term Memory, and Gated Recurrent Unit, evaluating each model's resilience to these sophisticated attacks. The experimental results suggest that the hybrid model has achieved a detection rate of up to 99.96%. The findings demonstrate the limitations of existing models and emphasize the need for enhanced defensive mechanisms, shedding light on future developments for securing sensitive data against evolving side-channel threats.