Adversarial domain adaptation is widely adopted for knowledge transfer in unsupervised medical imaging scenarios. However, existing methods often suffer from rigid alignment strategies, limiting their efficacy in complex clinical dataset shifts caused by heterogeneous scanners. To address this, we propose a bi-directional mutually-supervised (BDMS) framework to find an optimal intermediate distribution that combines the prior distributions of both domains. First, feature extractors are established for both source and target domains to fit their respective posterior distributions. Then, through mutual supervision, a multiple bidirectional state is reached. In order to ensure improved classification performance and the acquisition of distinct information in each domain, we propose the utilization of mutually learning classifiers. These classifiers are designed to stabilize dynamic changes during training and expedite convergence. Additionally, we introduce a novel conditional consistency loss to regulate the conditional distributions and further enhance the overall performance. Extensive experiments on multi-center medical datasets (breast ultrasound and endoscopic ultrasound) and natural image benchmarks (Office-31) demonstrate that BDMS achieves state-of-the-art performance, validates the universality and robustness of our approach.

With the advent of big data, modern businesses face an increasing need to store and process large volumes of sensitive customer information on the cloud. In these environments, resources are shared across a multitude of mutually untrusting tenants increasing propensity for data leakage. This problem stands to grow further in severity with increasing use of clouds in all aspects of our daily lives and the recent spate of high-profile data exfiltration attacks are evidence. To highlight this serious issue, we present a novel and highspeed network-based covert channel that is robust and circumvents a broad set of security mechanisms currently deployed by cloud vendors. We successfully test our channel on numerous network environments, including commercial clouds such as EC2 and Azure. Using an information theoretic model of the channel, we derive an upper bound on the maximum information rate and propose an optimal coding scheme. Our adaptive decoding algorithm caters to the cross traffic in the channel and maintains high bit rates and extremely low error rates. Finally, we discuss several effective avenues for mitigation of the aforementioned channel and provide insights into how data exfiltration can be prevented in such shared environments.

When multiple job processes are served by a single scheduler, the queueing delays of one process are often affected by the others, resulting in a timing side channel that leaks the arrival pattern of one process to the others. In this work, we study such a timing side channel between a regular user and a malicious attacker. Utilizing Shannon’s mutual information as a measure of information leakage between the user and attacker, we analyze privacy-preserving behaviors of common work-conserving schedulers. We find that the attacker can always learn perfectly the user’s arrival process in a longest-queue-first (LQF) scheduler. When the user’s job arrival rate is very low (near zero), first-come–first-serve (FCFS) and round-robin schedulers both completely reveal the user’s arrival pattern. The near-complete information leakage in the low-rate traffic region is proven to be reduced by half in a work-conserving version of TDMA (WC-TDMA) scheduler, which turns out to be privacy-optimal in the class of deterministic working-conserving (det-WC) schedulers, according to a universal lower bound on information leakage we derive for all det-WC schedulers.

The capacity of a queueing timing channel in which a user modulates messages to another user via his pattern of access to a shared resource scheduled in an FCFS manner is calculated. One example of such a channel is the cross-Virtual Network (VN) covert channel in data center networks. In data center networks, software-defined-networks generate logically isolated virtual networks, across which direct data exchange is impossible. However, since packet flows belonging to different VNs inevitably share underlying network infrastructure, it is possible to transfer data across VNs through timing channels resulting from the queueing effects of the shared resource.

Tor is a widely used network for anonymous communication. Its users frequently experience large communication delays, due to the high user-to-relay ratio, the bandwidth-intensive BitTorrent transfers of a small fraction of the user base, and the inherent latencies from routing traffic through multiple relay hops scattered around the world. These delays significantly degrade the user experience of web browsing, a dominant use of Tor.

Traffic analysis often requires direct observations of network connections at local vantage points. In this work, we show that traffic analysis can be performed remotely by taking advantage of a timing side channel. The timing side channel results from a shared resource, namely, the scheduler between two users. Utilizing Shannon equivocation as a privacy metric, we prove that one user can learn the complete traffic pattern of the other user if the scheduler employs a first come first serve (FCFS) policy. Moreover, we show the feasibility of a real system attack exploiting the timing side channel inside a home digital subscriber line (DSL) router. This demonstrates the magnitude of the threat timing side channels pose for traffic analysis.

Flow watermarks efficiently link packet flows in a network in order to thwart various attacks such as stepping stones. We study the problem of designing good flow watermarks. Earlier flow watermarking schemes mostly considered substitution errors, neglecting the effects of packet insertions and deletions that commonly happen within a network. More recent schemes considered packet deletions but often at the expense of the watermark visibility. We present an invisible flow watermarking scheme capable of enduring a large number of packet losses and insertions. To maintain invisibility, our scheme uses quantization-index-modulation (QIM) to embed the watermark into interpacket delays, as opposed to time intervals including many packets. As the watermark is injected within individual packets, packet losses and insertions may lead to watermark desynchronization and substitution errors. To address this issue, we add a layer of error-correction coding to our scheme. Experimental results on both synthetic and real network traces demonstrate that our scheme is robust to network jitter, packet drops, and splits, while remaining invisible to an attacker.

A key challenge in censorship-resistant web browsing is being able to direct legitimate users to redirection proxies while preventing censors, posing as insiders, from discovering their addresses and blocking them. We propose a new framework for censorship-resistant web browsing called CensorSpoofer that addresses this challenge by exploiting the asymmetric nature of web browsing traffic and making use of IP spoofing. CensorSpoofer de-couples the upstream and downstream channels, using a low-bandwidth indirect channel for delivering outbound requests (URLs) and a high-bandwidth direct channel for downloading web content. The upstream channel hides the request contents using steganographic encoding within Email or instant messages, whereas the downstream channel uses IP address spoofing so that the real address of the proxies is not revealed either to legitimate users or censors. We built a proof-of-concept prototype that uses encrypted VoIP for this downstream channel and demonstrated the feasibility of using the CensorSpoofer framework in a realistic environment.

Flow watermarking1is an efficient technique for linking packet flows that helps thwart various attacks in networks such as over the Internet. Current state-of-the-art water-marking schemes withstand packet losses at the expense of compromising invisibility. We present an invisible flow watermarking scheme that can endure large numbers of packet losses. To maintain invisibility, our scheme embeds quantization-index modulation watermarks into inter-packet delays (as opposed to intervals). As the watermark is injected within individual packets, packet losses may lead to water-mark desynchronization and substitution errors. To deal with this issue we propose a maximum likelihood decoding (ML) scheme based on a hidden-Markov model (HMM) of the channel. Experimental results demonstrate that our scheme is robust to both network jitters and packet deletions while remaining invisible to an attacker.

The information leakage of a queuing side channel in two-user-shared scheduling system is studied from an information theoretic perspective. In the queueing side channel, a malicious attacker can learn the pattern of jobs from a legitimate user using the queuing delays experienced at the shared buffer. An analytical framework is proposed to quantify information leakage using Shannon's equivocation, and the information leakage of the standard First-come-First-serve scheduler is studied in a slotted system with geometric arrivals. The analysis of the FCFS scheduler demonstrates that the policy provides “good privacy” when arrival rates are very low; the leaked information increases with the rate of the attacker's jobs and approaches the maximum retrievable information as the sum-rate of arrivals approaches the boundary of the stability region of the queue.

Recent work has shown that traffic analysis of data carried on encrypted tunnels can be used to recover important semantic information. As one example, attackers can find out which website, or which page on a website, a user is accessing simply by monitoring the traffic patterns. We show that traffic analysis is a much greater threat to privacy than previously thought, as such attacks can be carried out remotely. In particular, we show that, to perform traffic analysis, adversaries do not need to directly observe the traffic patterns. Instead, they can send probes from a far-off vantage point that exploit a queuing side channel in routers.

We examine a queuing side channel which results from a shared resource between two users in the context of packet networks. We consider the scenario where one of them is a legitimate user and the other is an attacker who is trying to learn about the former's activities. We show that the waiting time of an adversary sending a small but frequent probe stream to the shared resource (e.g., a router) is highly correlated with traffic pattern of the user.

This paper presents a dangerous low-cost traffic analysis attack in packet-based networks, such as the Internet. The attack is mountable in any scenario where a shared routing resource exists among users. A real-world attack successfully compromised the privacy of a user without requiring significant resources in terms of access, memory, or computational power. The effectiveness of our attack is demonstrated in a scenario where the user's DSL router uses FCFS scheduling policy. Specifically, we show that by using a low-rate sequence of probes, a remote attacker can obtain significant traffic-timing and volume information about a particular user, just by observing the round trip time of the probes. We also observe that even when the scheduling policy is changed to round-robin, while the correlation reduces significantly, the attacker can still reliably deduce user's traffic pattern. Most of the router scheduling policies designed to date are evaluated mostly on the metrics of throughput, delay and fairness. Our work is aimed to demonstrate a need for considering an additional metric that quantifies the information leak between the individual traffic flows through the router.