This paper introduces an invisible attack called Infrared Laser Reflection (ILR) that exploits the sensitivity of IR-sensitive cameras to mislead traffic sign recognition systems in connected autonomous vehicles (CAVs). Unlike previous attacks that rely on visible stickers or projected patterns, ILR uses infrared (IR) laser reflections that are invisible to humans but visible to IR-sensitive cameras. The attack leverages the properties of IR laser reflections to induce misclassification of traffic signs, such as stop signs and speed limit signs, by altering the camera's output images. The attack is designed to be stealthy and effective in both indoor and outdoor scenarios, with high success rates in real-world conditions. The ILR attack works by projecting an IR laser onto a target traffic sign, causing the camera to capture the reflected IR light as part of the image. This altered image is then misclassified by the vehicle's perception module, leading to dangerous misinterpretations. The attack is effective because it exploits the lack of IR filters in some CAV cameras, which allows the IR light to be captured and processed by the camera. The attack is optimized using a black-box optimization methodology that automatically finds the optimal location and size of the IR pattern to maximize the attack's effectiveness while minimizing the required laser power. The paper evaluates the effectiveness of the ILR attack against two major traffic sign recognition architectures using images captured with four different IR-sensitive cameras. The results show that the attack achieves a 100% success rate in indoor laboratory conditions and a 80.5% success rate in outdoor scenarios with moving vehicles. The attack is also evaluated against the current state-of-the-art certifiable defense, PatchCleanser, which is found to be ineffective against ILR attacks as it mis-certifies a significant portion of cases. The paper also proposes a detection strategy based on the physical properties of IR laser reflections, which can detect 96% of ILR attacks. The study highlights the limitations of existing defenses against IR-based attacks and demonstrates the potential of ILR as a novel, stealthy attack vector for traffic sign recognition systems. The results show that the ILR attack is highly effective in real-world scenarios and can be used to mislead CAV perception systems, emphasizing the need for robust defenses against such attacks.This paper introduces an invisible attack called Infrared Laser Reflection (ILR) that exploits the sensitivity of IR-sensitive cameras to mislead traffic sign recognition systems in connected autonomous vehicles (CAVs). Unlike previous attacks that rely on visible stickers or projected patterns, ILR uses infrared (IR) laser reflections that are invisible to humans but visible to IR-sensitive cameras. The attack leverages the properties of IR laser reflections to induce misclassification of traffic signs, such as stop signs and speed limit signs, by altering the camera's output images. The attack is designed to be stealthy and effective in both indoor and outdoor scenarios, with high success rates in real-world conditions. The ILR attack works by projecting an IR laser onto a target traffic sign, causing the camera to capture the reflected IR light as part of the image. This altered image is then misclassified by the vehicle's perception module, leading to dangerous misinterpretations. The attack is effective because it exploits the lack of IR filters in some CAV cameras, which allows the IR light to be captured and processed by the camera. The attack is optimized using a black-box optimization methodology that automatically finds the optimal location and size of the IR pattern to maximize the attack's effectiveness while minimizing the required laser power. The paper evaluates the effectiveness of the ILR attack against two major traffic sign recognition architectures using images captured with four different IR-sensitive cameras. The results show that the attack achieves a 100% success rate in indoor laboratory conditions and a 80.5% success rate in outdoor scenarios with moving vehicles. The attack is also evaluated against the current state-of-the-art certifiable defense, PatchCleanser, which is found to be ineffective against ILR attacks as it mis-certifies a significant portion of cases. The paper also proposes a detection strategy based on the physical properties of IR laser reflections, which can detect 96% of ILR attacks. The study highlights the limitations of existing defenses against IR-based attacks and demonstrates the potential of ILR as a novel, stealthy attack vector for traffic sign recognition systems. The results show that the ILR attack is highly effective in real-world scenarios and can be used to mislead CAV perception systems, emphasizing the need for robust defenses against such attacks.