arXiv subjects

Ziqi Xu

Publications and source records attributed to Ziqi Xu.

76 records · Page 5Linked to original sources

arXiv2022

Causal Inference with Conditional Instruments using Deep Generative Models

The instrumental variable (IV) approach is a widely used way to estimate the causal effects of a treatment on an outcome of interest from observational data with latent confounders. A standard IV is expected to be related to the treatment variable and independent of all other variables in the system. However, it is challenging to search for a standard IV from data directly due to the strict conditions. The conditional IV (CIV) method has been proposed to allow a variable to be an instrument conditioning on a set of variables, allowing a wider choice of possible IVs and enabling broader practical applications of the IV approach. Nevertheless, there is not a data-driven method to discover a CIV and its conditioning set directly from data. To fill this gap, in this paper, we propose to learn the representations of the information of a CIV and its conditioning set from data with latent confounders for average causal effect estimation. By taking advantage of deep generative models, we develop a novel data-driven approach for simultaneously learning the representation of a CIV from measured variables and generating the representation of its conditioning set given measured variables. Extensive experiments on synthetic and real-world datasets show that our method outperforms the existing IV methods.

cs.LG↗

arXiv2022

Wiggle: Physical Challenge-Response Verification of Vehicle Platooning

Autonomous vehicle platooning promises many benefits such as fuel efficiency, road safety, reduced traffic congestion, and passenger comfort. Platooning vehicles travel in a single file, in close distance, and at the same velocity. The platoon formation is autonomously maintained by a Cooperative Adaptive Cruise Control (CACC) system which relies on sensory data and vehicle-to-vehicle (V2V) communications. In fact, V2V messages play a critical role in shortening the platooning distance while maintaining safety. Whereas V2V message integrity and source authentication can be verified via cryptographic methods, establishing the truthfulness of the message contents is a much harder task. This work establishes a physical access control mechanism to restrict V2V messages to platooning members. Specifically, we aim at tying the digital identity of a candidate requesting to join a platoon to its physical trajectory relative to the platoon. We propose the {\em Wiggle} protocol that employs a physical challenge-response exchange to prove that a candidate requesting to be admitted into a platoon actually follows it. The protocol name is inspired by the random longitudinal movements that the candidate is challenged to execute. {\em Wiggle} prevents any remote adversary from joining the platoon and injecting fake CACC messages. Compared to prior works, {\em Wiggle} is resistant to pre-recording attacks and can verify that the candidate is directly behind the verifier at the same lane.

cs.CR↗

arXiv2022

Assessing Classifier Fairness with Collider Bias

The increasing application of machine learning techniques in everyday decision-making processes has brought concerns about the fairness of algorithmic decision-making. This paper concerns the problem of collider bias which produces spurious associations in fairness assessment and develops theorems to guide fairness assessment avoiding the collider bias. We consider a real-world application of auditing a trained classifier by an audit agency. We propose an unbiased assessment algorithm by utilising the developed theorems to reduce collider biases in the assessment. Experiments and simulations show the proposed algorithm reduces collider biases significantly in the assessment and is promising in auditing trained classifiers.

cs.LG↗

arXiv2021

Man-in-the-Middle Attack Resistant Secret Key Generation via Channel Randomization

Physical-layer based key generation schemes exploit the channel reciprocity for secret key extraction, which can achieve information-theoretic secrecy against eavesdroppers. Such methods, although practical, have been shown to be vulnerable against man-in-the-middle (MitM) attacks, where an active adversary, Mallory, can influence and infer part of the secret key generated between Alice and Bob by injecting her own packet upon observing highly correlated channel/RSS measurements from Alice and Bob. As all the channels remain stable within the channel coherence time, Mallory's injected packets cause Alice and Bob to measure similar RSS, which allows Mallory to successfully predict the derived key bits. To defend against such a MitM attack, we propose to utilize a reconfigurable antenna at one of the legitimate transceivers to proactively randomize the channel state across different channel probing rounds. The randomization of the antenna mode at every probing round breaks the temporal correlation of the channels from the adversary to the legitimate devices, while preserving the reciprocity of the channel between the latter. This prevents key injection from the adversary without affecting Alice and Bob's ability to measure common randomness. We theoretically analyze the security of the protocol and conduct extensive simulations and real-world experiments to evaluate its performance. Our results show that our approach eliminates the advantage of an active MitM attack by driving down the probability of successfully guessing bits of the secret key to a random guess.

cs.CR↗