Jianhua Wang

Junchao Fan, Qi Wei, Ruichen Zhang et al.

Deep reinforcement learning (DRL) has demonstrated remarkable success in developing autonomous driving policies. However, its vulnerability to adversarial attacks remains a critical barrier to real-world deployment. Although existing robust methods have achieved success, they still suffer from three key issues: (i) these methods are trained against myopic adversarial attacks, limiting their abilities to respond to more strategic threats, (ii) they have trouble causing truly safety-critical events (e.g., collisions), but instead often result in minor consequences, and (iii) these methods can introduce learning instability and policy drift during training due to the lack of robust constraints. To address these issues, we propose Intelligent General-sum Constrained Adversarial Reinforcement Learning (IGCARL), a novel robust autonomous driving approach that consists of a strategic targeted adversary and a robust driving agent. The strategic targeted adversary is designed to leverage the temporal decision-making capabilities of DRL to execute strategically coordinated multi-step attacks. In addition, it explicitly focuses on inducing safety-critical events by adopting a general-sum objective. The robust driving agent learns by interacting with the adversary to develop a robust autonomous driving policy against adversarial attacks. To ensure stable learning in adversarial environments and to mitigate policy drift caused by attacks, the agent is optimized under a constrained formulation. Extensive experiments show that IGCARL improves the success rate by at least 27.9% over state-of-the-art methods, demonstrating superior robustness to adversarial attacks and enhancing the safety and reliability of DRL-based autonomous driving.

Jianhua Wang, Lu Qin, Baofeng Wu

In this paper, we improve the cube attack by exploiting low-degree factors of the superpoly w.r.t. certain "special" index set of cube (ISoC). This can be viewed as a special case of the correlation cube attack proposed at Eurocrypt 2018, but under our framework more beneficial equations on the key variables can be obtained in the key-recovery phase. To mount our attack, one has two challenging problems: effectively recover algebraic normal form of the superpoly and extract out its low-degree factors; and efficiently search a large quantity of good ISoCs. We bring in new techniques to solve both of them. First, we propose the variable substitution technique for middle rounds of a cipher, in which polynomials on the key variables in the algebraic expressions of internal states are substituted by new variables. This will improve computational complexity of the superpoly recovery and promise more compact superpolys that can be easily decomposed with respect to the new variables. Second, we propose the vector numeric mapping technique, which seeks out a tradeoff between efficiency of the numeric mapping technique and accuracy of the monomial prediction technique in degree evaluation of superpolys. Combining with this technique, a fast pruning method is given and modeled by MILP to filter good ISoCs of which the algebraic degree satisfies some fixed threshold. Thanks to automated MILP solvers, it becomes practical to comprehensively search for good cubes across the entire search space. To illustrate the power of our techniques, we apply all of them to Trivium stream cipher. The previous best practical key recovery attack was on 820-round Trivium with complexity $2^{53.17}$. We put forward 820-, 825- and 830-round practical key-recovery attacks, in which there are 2^{80}\times 87.8%, 2^{80}\times 83% and 2^{80}\times 65.7% keys that could be practically recovered, respectively.

Yixiang Wang, Jiqiang Liu, Xiaolin Chang et al.

White-box Adversarial Example (AE) attacks towards Deep Neural Networks (DNNs) have a more powerful destructive capacity than black-box AE attacks in the fields of AE strategies. However, almost all the white-box approaches lack interpretation from the point of view of DNNs. That is, adversaries did not investigate the attacks from the perspective of interpretable features, and few of these approaches considered what features the DNN actually learns. In this paper, we propose an interpretable white-box AE attack approach, DI-AA, which explores the application of the interpretable approach of the deep Taylor decomposition in the selection of the most contributing features and adopts the Lagrangian relaxation optimization of the logit output and L_p norm to further decrease the perturbation. We compare DI-AA with six baseline attacks (including the state-of-the-art attack AutoAttack) on three datasets. Experimental results reveal that our proposed approach can 1) attack non-robust models with comparatively low perturbation, where the perturbation is closer to or lower than the AutoAttack approach; 2) break the TRADES adversarial training models with the highest success rate; 3) the generated AE can reduce the robust accuracy of the robust black-box models by 16% to 31% in the black-box transfer attack.