Jianting Zhang

Zhongtang Luo, Jianting Zhang, Akshat Neerati et al.

The Tor network offers network anonymity to its users by routing their traffic through a sequence of relays. A group of nine directory authorities maintains information about all available relay nodes using a distributed directory protocol. We observe that the current protocol makes a strong synchrony assumption, which makes it vulnerable to natural as well as adversarial non-synchronous communication scenarios over the Internet. In this paper, we show that it is possible to cause a failure in the Tor directory protocol by targeting a majority of the authorities for only five minutes using a well-executed distributed denial-of-service (DDoS) attack. We demonstrate this attack in a controlled environment and show that it is cost-effective for as little as \$53.28 per month to disrupt the protocol and to effectively bring down the entire Tor network. To mitigate this problem, we consider the popular partial synchrony assumption that ensures protocol security even when the network delays are large and unknown initially. We design a new Tor directory protocol that leverages a standard partial-synchronous consensus protocol to solve this problem, while also proving its security. We have implemented a prototype in Rust, demonstrating comparable performance to the current protocol.

Xinwei Liu, Junyuan Liang, Zicong Hong et al.

Augmenting model-free reinforcement learning (RL) with representations learned through observation dynamics prediction (observation-predictive RL) can improve sample efficiency and performance, with minor modifications and limited additional computation. However, this approach still struggles in challenging tasks with low-dimensional observations. In this paper, we identify a key factor behind this problem: unbalanced reconstruction losses across observation dimensions, where dimensions with larger value ranges dominate the loss. This encourages the agent to neglect dimensions with relatively small ranges, leading to degraded performance. To address this issue, we propose a novel normalization method tailored to online RL, which normalizes low-dimensional observations and balances the resulting losses and gradients. Beyond balancing reconstruction losses, observation normalization enables dynamics prediction to be performed in a normalized observation space, thereby providing a unified treatment of low- and high-dimensional inputs (e.g., physical states and images). Building on this idea, we further introduce Normalized Observation Space Dynamics-Augmented Q-learning (NASDAQ), a framework for observation-predictive RL applicable across diverse domains. NASDAQ learns state-action representations by coupling value learning with two auxiliary tasks: short-term value prediction and next normalized observation prediction. Extensive experiments demonstrate that NASDAQ achieves competitive or superior performance compared with state-of-the-art model-based and self-predictive RL methods, while requiring significantly less training wall-time.