Total: 85
Organizations that provide essential services such as electricity, healthcare, and secure financial transactions are required to use digital-security compliance programs to establish a baseline of minimum security. Unfortunately, these compliance programs are known to suffer from a multitude of issues (both in how they are written and in how organizations implement them), resulting in organizations implementing their own security measures to fill actual or perceived compliance gaps. In this study, we survey 40 security professionals from six U.S. essential-service sectors to gain insight into how organizations complement compliance to fix perceived security gaps, which measures worked particularly well, and how their organizations prioritize and evaluate the measures they adopt. We find that organizations complement compliance programs often, with 37 of 40 participants confirming that their organizations have gone beyond what they perceive as mandated compliance measures to mitigate otherwise unaddressed risks. While participants were generally positive about these perceived complementary measures, they also reported challenges related to poor management, information saturation, and difficulty keeping complementary measures up-to-date and relevant. Based on these results, we recommend that compliance standards directly integrate guidance for carefully managing and auditing any perceived complementary measures that an organization chooses to implement and that organizations carefully plan end-to-end deployment and operation before implementing these measures.
Binary search is one of the most popular algorithms in computer science. Realizing it in the context of secure multiparty computation which demands data-oblivious execution, however, is extremely non-trivial. It has been previously implemented only using oblivious RAM (ORAM) for secure computation and in this work we initiate the study of this topic using conventional secure computation techniques based on secret sharing. We develop a suite of protocols with different properties and of different structure for searching a private dataset of $m$ elements by a private numeric key. Our protocols result in $O(m)$ and $O(sqrt{m})$ communication using only standard and readily available operations based on secret sharing. We further extend our protocols to support write operations, namely, binary search that obliviously updates the selected element, and realize two variants: updating non-key fields and updating the key field. Our implementation results indicate that even after applying known and our own optimizations to the fastest ORAM constructions, our solutions are faster than optimized ORAM schemes for datasets of up to $2^{30}$ elements and by up to 2 orders of magnitude. We hope that this work will prompt further interest in seeking efficient realizations of this important problem.
Software vulnerabilities are one of the major threats to computer security and have caused substantial damage over the past decades. Consequently, numerous techniques have been proposed to mitigate the risk of exploitation of vulnerable programs. One of the most relevant defense mechanisms is Control-Flow Integrity (CFI): multiple variants have been introduced and extensively discussed in academia as well as deployed in the industry. However, it is hard to compare the security guarantees of these implementations as existing metrics (such as AIR) do not consider the different usefulness to the attacker of different basic blocks, which are the fundamental components that constitute the code of any application. This paper introduces BlockInsulation and CFGInsulation, novel metrics designed to overcome this limitation by modeling the usefulness of basic blocks for an attacker trying to traverse the program’s control-flow graph. Moreover, we propose a new CFI policy generator, named NumCFI, which is orthogonal to existing policy generators and prevents the attacker from taking shortcuts from vulnerable code to a system call instruction. We evaluate NumCFI, as well as a number of other CFI policy generators, using BlockInsulation, CFGInsulation, and existing metrics. Lastly, we describe l+tCFI, our implementation that combines NumCFI and an existing label-based policy, with a performance overhead of just 1.27%.
Shared cache resources in multi-core processors are vulnerable to cache side-channel attacks. Recently proposed defenses such as randomized mapping of addresses to cache lines or well-known cache partitioning have their own caveats: Randomization-based defenses have been shown vulnerable to newer attack algorithms besides relying on weak cryptographic primitives. They do not fundamentally address the root cause for cache side-channel attacks, namely, mutually distrusting codes sharing cache resources. Cache partitioning defenses provide the strict resource partitioning required to effectively block all side-channel threats. However, they usually rely on way-based partitioning which is not fine-grained and cannot scale to support a larger number of protection domains, e.g., in trusted execution environment (TEE) security architectures, besides degrading performance and often resulting in cache underutilization. To overcome the shortcomings of both approaches, we present a novel and flexible set-associative cache partitioning design for TEE architectures, called Chunked-Cache. The core idea of Chunked-Cache is to enable an execution context to “carve” out an exclusive configurable chunk of the cache if the execution requires side-channel resilience. If side-channel resilience is not required, mainstream cache resources can be freely utilized. Hence, our proposed cache design addresses the security performance trade-off practically by enabling efficient selective and on-demand utilization of side-channel-resilient caches, while providing well-grounded future-proof security guarantees. We show that Chunked-Cache provides side-channel-resilient cache utilization for sensitive code execution, with small hardware overhead, while incurring no performance overhead on the OS. We also show that it outperforms conventional way-based cache partitioning by 43%, while scaling significantly better to support a larger number of protection domains.
Federated Learning (FL) allows multiple clients to collaboratively train a Neural Network (NN) model on their private data without revealing the data. Recently, several targeted poisoning attacks against FL have been introduced. These attacks inject a backdoor into the resulting model that allows adversary-controlled inputs to be misclassified. Existing countermeasures against backdoor attacks are inefficient and often merely aim to exclude deviating models from the aggregation. However, this approach also removes benign models of clients with deviating data distributions, causing the aggregated model to perform poorly for such clients. To address this problem, we propose emph{DeepSight}, a novel model filtering approach for mitigating backdoor attacks. It is based on three novel techniques that allow to characterize the distribution of data used to train model updates and seek to measure fine-grained differences in the internal structure and outputs of NNs. Using these techniques, DeepSight can identify suspicious model updates. We also develop a scheme that can accurately cluster model updates. Combining the results of both components, DeepSight is able to identify and eliminate model clusters containing poisoned models with high attack impact. We also show that the backdoor contributions of possibly undetected poisoned models can be effectively mitigated with existing weight clipping-based defenses. We evaluate the performance and effectiveness of DeepSight and show that it can mitigate state-of-the-art backdoor attacks with a negligible impact on the model's performance on benign data.
Mutation-based fuzzing is one of the most popular approaches to discover vulnerabilities in a program. To alleviate the inefficiency of mutation-based fuzzing incurred by high randomness in the mutation process, multiple solutions are developed in recent years, especially coverage-based fuzzing. They mainly employ adaptive mutation strategies or integrate constraint-solving techniques to make a good exploration of the test cases which trigger unique paths and crashes. However, they lack a fine-grained reusing of fuzzing history to construct these interesting test cases, i.e., they largely fail to properly utilize fuzzing history across different fuzzing trials. In fact, we discover that test cases in fuzzing history contain rich knowledge of the key mutation strategies that lead to the discovery of unique paths and crashes. Specifically, partial path constraint solutions implicitly carried in these mutation strategies can be reused to accelerate the discovery of new paths and crashes that share similar partial path constraints. Therefore, we first propose a lightweight and efficient Probabilistic Byte Orientation Model (PBOM) that properly captures the byte-level mutation strategies from intra- and inter-trial history and thus can effectively trigger unique paths and crashes. We then present a novel history-driven mutation framework named EMS that employs PBOM as one of the mutation operators to probabilistically provide desired mutation byte values according to the input ones. We evaluate EMS against state-of-the-art fuzzers including AFL, QSYM, MOPT, MOPT-dict, EcoFuzz, and AFL++ on 9 real world programs. The results show that EMS discovers up to 4.91X more unique vulnerabilities than the baseline, and finds more line coverage than other fuzzers on most programs. We report all of the discovered new vulnerabilities to vendors and will open source the prototype of EMS on GitHub.
Path tracing tools, such as traceroute, are simple yet fundamental network debugging tools for network operators to detect and fix network failures. Unfortunately, adversaries can also use such tools to retrieve previously unknown network topology information which is key to realizing sophisticated Denial-of-Service attacks, such as Link Flooding Attacks (LFAs), more efficiently. Over the last few years, several network obfuscation defenses have been proposed to proactively mitigate LFAs by exposing virtual (fake) topologies that conceal potential bottleneck network links from adversaries. However, to date there has been no comprehensive and systematic analysis of the level of security and utility their virtual topologies offer. A critical analysis is thus a necessary step towards better understanding their limitations and building stronger and more practical defenses against LFAs. In this paper, we first conduct a security analysis of the three state-of-the-art network obfuscation defenses. Our analysis reveals four important, common limitations that can significantly decrease the security and utility of their virtual topologies. Motivated by our findings, we present EqualNet, a secure and practical proactive defense for long-term network topology obfuscation that alleviates LFAs within a network domain. EqualNet aims to equalize tracing flow distributions over nodes and links so that adversaries are unable to distinguish which of them are the most important ones, thus significantly increasing the cost of performing LFAs. Meanwhile, EqualNet preserves subnet information, helping network operators who use path tracing tools to debug their networks. To demonstrate its feasibility, we implement a full prototype of it using Software-Defined Networking (SDN) and perform extensive evaluations both in software and hardware. Our results show that EqualNet is effective at equalizing the tracing flow distributions of small, medium and large networks even when only a small number of routers within the network support SDN. Finally, we analyze the security of EqualNet against a wide variety of attacks.
In the present era of ubiquitous digitization more and more services are becoming available online which is amplified by the Corona pandemic. The fast-growing mobile service market opens up new attack surfaces to the mobile service ecosystem. Hence, mobile service providers are faced with various challenges to protect their services and in particular the associated mobile apps. Defenses for apps are, however, often limited to (lightweight) application-level protection such as app hardening and monitoring and intrusion detection. Therefore, effective risk management is crucial to limit the exposure of mobile services to threats and potential damages caused by attacks. In this paper, we present FedCRI, a solution for sharing Cyber-Risk Intelligence (CRI). At its core, FedCRI transforms mobile cyber-risks into machine learning (ML) models and leverages ML-based risk management to evaluate security risks on mobile devices. FedCRI enables fast and autonomous sharing of actionable ML-based CRI knowledge by utilizing Federated Learning (FL). FL allows collaborative training of effective risk detection models based on information contributed by different mobile service providers while preserving the privacy of the training data of the individual organizations. We extensively evaluate our approach on several real-world user databases representing 23.8 million users of security-critical mobile apps (since Android 4 and iOS 6) provided by nine different service providers in different European countries. The datasets were collected over the course of six years in the domains of financial services, payments, or insurances. Our approach can successfully extract accurate CRI models, allowing the effective identification of cybersecurity risks on mobile devices. Our evaluation shows that the federated risk detection model can achieve better than 99% accuracy in terms of F1-score in most risk classification tasks with a very low number of false positives.
Recent years have seen a strong uptick in both the prevalence and real-world consequences of false information spread through online platforms. At the same time, encrypted messaging systems such as WhatsApp, Signal, and Telegram, are rapidly gaining popularity as users seek increased privacy in their digital lives. The challenge we address is how to combat the viral spread of misinformation without compromising privacy. Our FACTS system tracks user complaints on messages obliviously, only revealing the message's contents and originator once sufficiently many complaints have been lodged. Our system is *private*, meaning it does not reveal anything about the senders or contents of messages which have received few or no complaints; *secure*, meaning there is no way for a malicious user to evade the system or gain an outsized impact over the complaint system; and *scalable*, as we demonstrate excellent practical efficiency for up to millions of complaints per day. Our main technical contribution is a new collaborative counting Bloom filter, a simple construction with difficult probabilistic analysis, which may have independent interest as a privacy-preserving randomized count sketch data structure. Compared to prior work on message flagging and tracing in end-to-end encrypted messaging, our novel contribution is the addition of a high threshold of multiple complaints that are needed before a message is audited or flagged. We present and carefully analyze the probabilistic performance of our data structure, provide a precise security definition and proof, and then measure the accuracy and scalability of our scheme via experimentation.
Smartphones today leverage baseband processors to implement the multitude of cellular protocols. Basebands execute firmware, which is responsible for decoding hundreds of message types developed from three decades of cellular standards. Despite its large over-the-air attack surface, baseband firmware has received little security analysis. Previous work mostly analyzed only a handful of firmware images from a few device models, but often relied heavily on time-consuming manual static analysis or single-function fuzzing. To fill this gap, we present FirmWire, the first full-system emulation platform for baseband processors that executes unmodified baseband binary firmware. FirmWire provides baseband-specific APIs to easily add support for new vendors, firmware images, and security analyses. To demonstrate FirmWire’s scalability, we support 213 firmware images across 2 vendors and 9 phone models, allowing them to be executed and tested. With these images, FirmWire automatically discovers and bridges internal baseband APIs, allowing protocol messages to be injected with ease. Using these entry points, we selected the LTE and GSM protocols for fuzzing and discovered 7 pre-authentication memory corruptions that could lead to remote code execution--4 of which were previously unknown. We reproduced these crashes over-the-air on real devices, proving FirmWire’s emulation accuracy. FirmWire is a scalable platform for baseband security testing and we release it as open-source to the community for future research.
Causality analysis is an effective technique for investigating and detecting cyber attacks. However, by focusing on auditing at the Operating System level, existing causal analysis techniques lack visibility into important application-level semantics, such as configuration changes that control application runtime behavior. This leads to incorrect attack attribution and half-baked tracebacks. In this work, we propose Dossier, a specialized provenance tracker that enhances the visibility of the Linux auditing infrastructure. By providing additional hooks into the system, Dossier can generate a holistic view of the target application’s event history and causal chains, particularly those pertaining to configuration changes that are among the most common attack vectors observed in the real world. The extra “vantage points” in Dossier enable forensic investigators to bridge the semantic gap and correctly piece together attack fragments. Dossier leverages the versatility of information flow tracking and system call introspection to track all configuration changes, including both dynamic modifications that are applied directly to configuration-related program variables in memory and revisions to configuration files on disk with negligible runtime overhead (less than 7%). Evaluation on realistic workloads and real-world attack scenarios shows that Dossier can effectively reason about configuration-based attacks and accurately reconstruct the whole attack stories.
Machine learning (ML) has established itself as a cornerstone for various critical applications ranging from autonomous driving to authentication systems. However, with this increasing adoption rate of machine learning models, multiple attacks have emerged. One class of such attacks is training time attack, whereby an adversary executes their attack before or during the machine learning model training. In this work, we propose a new training time attack against computer vision based machine learning models, namely model hijacking attack. The adversary aims to hijack a target model to execute a different task than its original one without the model owner noticing. Model hijacking can cause accountability and security risks since a hijacked model owner can be framed for having their model offering illegal or unethical services. Model hijacking attacks are launched in the same way as existing data poisoning attacks. However, one requirement of the model hijacking attack is to be stealthy, i.e., the data samples used to hijack the target model should look similar to the model's original training dataset. To this end, we propose two different model hijacking attacks, namely Chameleon and Adverse Chameleon, based on a novel encoder-decoder style ML model, namely the Camouflager. Our evaluation shows that both of our model hijacking attacks achieve a high attack success rate, with a negligible drop in model utility.
Online behavioral advertising, and the associated tracking paraphernalia, poses a real privacy threat. Unfortunately, existing privacy-enhancing tools are not always effective against online advertising and tracking. We propose HARPO, a principled learning-based approach to subvert online behavioral advertising through obfuscation. HARPO uses reinforcement learning to adaptively interleave real page visits with fake pages to distort a tracker’s view of a user’s browsing profile. We evaluate HARPO against real-world user profiling and ad targeting models used for online behavioral advertising. The results show that HARPO improves privacy by triggering more than 40% incorrect interest segments and 6×higher bid values. HARPO outperforms existing obfuscation tools by as much as 16×for the same overhead. HARPO is also able to achieve better stealthiness to adversarial detection than existing obfuscation tools. HARPO meaningfully advances the state-of-the-art in leveraging obfuscation to subvert online behavioral advertising.
Despite significant recent progress toward making multi-party computation (MPC) practical, no existing MPC library offers complete robustness---meaning guaranteed output delivery, including in the offline phase---in a network that even has intermittent delays. Importantly, several theoretical MPC constructions already ensure robustness in this setting. We observe that the key reason for this gap between theory and practice is the absence of efficient verifiable/complete secret sharing (VSS/CSS) constructions; existing CSS protocols either require a) challenging broadcast channels in practice or b) introducing computation and communication overhead that is at least quadratic in the number of players. This work presents hbACSS, a suite of optimal-resilience asynchronous complete secret sharing protocols that are (quasi)linear in both computation and communication overhead. Towards developing hbACSS, we develop hbPolyCommit, an efficient polynomial commitment scheme that is (quasi)linear (in the polynomial degree) in terms of computation and communication overhead without requiring a trusted setup. We implement our hbACSS protocols, extensively analyze their practicality, and observe that our protocols scale well with an increasing number of parties. In particular, we use hbACSS to generate MPC input masks: a useful primitive which had previously only been calculated nonrobustly in practice.
Threat detection and forensics have become an imperative objective for any digital forensic triage. Supervised approaches have been proposed for inferring system and network anomalies; including anomaly detection contributions using syslogs. Nevertheless, most works downplay the importance of the interpretability of a model's decision-making process. In this research, we are among the first to propose an interpretable federated transformer log learning model for threat detection supporting explainable cyber forensics. The proposed model is generated by training a local transformer-based threat detection model at each client in an organizational unit. Local models learn the system's normal behavior from the syslogs which keep records of execution flows. Subsequently, a federated learning server aggregates the learned model parameters from local models to generate a global federated learning model. Log time-series capturing normal behavior are expected to differ from those possessing cyber threat activity. We demonstrate this difference through a goodness of fit test based on Karl-Pearson's Chi-square statistic. To provide insights on actions triggering this difference, we integrate an attention-based interpretability module. We implement and evaluate our proposed model using HDFS, a publicly available log dataset, and an in-house collected and publicly-released dataset named CTDD, which consists of more than 8 million syslogs representing cloud collaboration services and systems compromised by different classes of cyber threats. Moreover, through different experiments, we demonstrate the log agnostic capability and applicability of our approach on real-world operational settings such as edge computing systems. Our interpretability module manifests significant attention difference between normal and abnormal logs which provide insightful interpretability of the model's decision-making process. Finally, we deem the obtained results as a validation for the appropriate adoption of our approach in achieving threat forensics in the real world.
Federated Learning (FL) allows multiple participants to train machine learning models collaboratively by keeping their datasets local while only exchanging model updates. Alas, this is not necessarily free from privacy and robustness vulnerabilities, e.g., via membership, property, and backdoor attacks. This paper investigates whether and to what extent one can use differential Privacy (DP) to protect both privacy and robustness in FL. To this end, we present a first-of-its-kind evaluation of Local and Central Differential Privacy (LDP/CDP) techniques in FL, assessing their feasibility and effectiveness. Our experiments show that both DP variants do defend against backdoor attacks, albeit with varying levels of protection-utility trade-offs, but anyway more effectively than other robustness defenses. DP also mitigates white-box membership inference attacks in FL, and our work is the first to show it empirically. Neither LDP nor CDP, however, defend against property inference. Overall, our work provides a comprehensive, re-usable measurement methodology to quantify the trade-offs between robustness/privacy and utility in differentially private FL.
Cooperative vehicle platooning significantly improves highway safety and fuel efficiency. In this model, a set of vehicles move in line formation and coordinate functions such as acceleration, braking, and steering using a combination of physical sensing and vehicle-to-vehicle (V2V) messaging. The authenticity and integrity of the V2V messages are paramount to highway safety. For this reason, recent V2V and V2X standards support the integration of a PKI. However, a PKI cannot bind a vehicle's digital identity to the vehicle's physical state (location, heading, velocity, etc.). As a result, a vehicle with valid cryptographic credentials can impact the platoon by creating ``ghost'' vehicles and injecting false state information. In this paper, we seek to provide the missing link between the physical and the digital world in the context of verifying a vehicle’s platoon membership. We focus on the property of following, where vehicles follow each other in a close and coordinated manner. We aim at developing a Proof-of-Following (PoF) protocol that enables a candidate vehicle to prove that it follows a verifier within the typical platooning distance. The main idea of the proposed {em PoF} protocol is to draw security from the common, but constantly changing environment experienced by the closely traveling vehicles. We use the large-scale fading effect of ambient RF signals as a common source of randomness to construct a {em PoF} primitive. The correlation of large-scale fading is an ideal candidate for the mobile outdoor environment because it exponentially decays with distance and time. We evaluate our PoF protocol on an experimental platoon of two vehicles in freeway, highway, and urban driving conditions. In such realistic conditions, we demonstrate that the PoF withstands both the pre-recording and following attacks with overwhelming probability.
Companies providing services like cab sharing, e-commerce logistics and, food delivery are willing to instrument their vehicles for scaling up measurements of traffic congestion, travel time, road surface quality, air quality, etc.~cite{polmeasure}. Analyzing fine-grained sensors data from such large fleets can be highly beneficial; however, this sensor information reveals the locations and the number of vehicles in the deployed fleet. This sensitive data is of high business value to rival companies in the same business domain, e.g., Uber vs. Ola, Uber vs. Lyft in cab sharing, or Amazon vs. Alibaba in the e-commerce domain. This paper provides privacy guarantees for the scenario mentioned above using Gaussian Process Regression (GPR) based interpolation, Differential Privacy (DP), and Secure two-party computations (2PC). The sensed values from instrumented vehicle fleets are made available preserving fleet and client privacy, along with client utility. Our system has efficient latency and bandwidth overheads, even for resource-constrained mobile clients. To demonstrate our end-to-end system, we build a sample Android application that gives the least polluted route alternatives given a source-destination pair in a privacy preserved manner.
While machine learning (ML) has made tremendous progress during the past decade, recent research has shown that ML models are vulnerable to various security and privacy attacks. So far, most of the attacks in this field focus on discriminative models, represented by classifiers. Meanwhile, little attention has been paid to the security and privacy risks of generative models, such as generative adversarial networks (GANs). In this paper, we propose the first set of training dataset property inference attacks against GANs. Concretely, the adversary aims to infer the macro-level training dataset property, i.e., the proportion of samples used to train a target GAN with respect to a certain attribute. A successful property inference attack can allow the adversary to gain extra knowledge of the target GAN's training dataset, thereby directly violating the intellectual property of the target model owner. Also, it can be used as a fairness auditor to check whether the target GAN is trained with a biased dataset. Besides, property inference can serve as a building block for other advanced attacks, such as membership inference. We propose a general attack pipeline that can be tailored to two attack scenarios, including the full black-box setting and partial black-box setting. For the latter, we introduce a novel optimization framework to increase the attack efficacy. Extensive experiments over four representative GAN models on five property inference tasks show that our attacks achieve strong performance. In addition, we show that our attacks can be used to enhance the performance of membership inference against GANs.
Network functions virtualization (NFV) enables agile deployment of network services on top of clouds. However, as NFV involves multiple levels of abstraction representing the same components, pinpointing the root cause of security incidents can become challenging. For instance, a security incident may be detected at a different level from where its root cause operations were conducted with no obvious link between the two. Moreover, existing provenance analysis techniques may produce results that are impractically large for human analysts to interpret due to the inherent complexity of NFV. In this paper, we propose ProvTalk, a provenance analysis system that handles the unique multi-level nature of NFV and assists the analyst to identify the root cause of security incidents. Specifically, we first define a multi-level provenance model to capture the dependencies between NFV levels. Next, we improve the interpretability through three novel techniques, i.e., multi-level pruning, mining-based aggregation, and rule-based natural language translation. We implement ProvTalk on a Tacker-OpenStack NFV platform and validate its effectiveness based on real-world security incidents. We demonstrate that ProvTalk captures management API calls issued to all NFV services, and produces more interpretable results by significantly reducing the size of the provenance graphs (up to 3.6 times smaller than the existing one-level pruning scheme and two orders of magnitude via multi-level aggregation scheme). Our user study shows that ProvTalk facilitates the analysis task of real-world users by generating more interpretable results.
Memory utilization can be reduced by merging identical memory blocks into copy-on-write mappings. Previous work showed that this so-called memory deduplication can be exploited in local attacks to break ASLR, spy on other programs, and determine the presence of data, i.e., website images. All these attacks exploit memory deduplication across security domains, which in turn was disabled. However, within a security domain or on an isolated system with no untrusted local access, memory deduplication is still not considered a security risk and was recently re-enabled on Windows by default. In this paper, we present the first fully remote memory-deduplication attacks. Unlike previous attacks, our attacks require no local code execution. Consequently, we can disclose memory contents from a remote server merely by sending and timing HTTP/1 and HTTP/2 network requests. We demonstrate our attacks on deduplication both on Windows and Linux and attack widely used server software such as Memcached and InnoDB. Our side channel leaks up to 34.41B/h over the internet, making it faster than comparable remote memory-disclosure channels. We showcase our remote memory-deduplication attack in three case studies: First, we show that an attacker can disclose the presence of data in memory on a server running Memcached. We show that this information disclosure channel can also be used for fingerprinting and detect the correct libc version over the internet in 166.51s. Second, in combination with InnoDB, we present an information disclosure attack to leak MariaDB database records. Third, we demonstrate a fully remote KASLR break in less than 4 minutes allowing to derandomize the kernel image of a virtual machine over the Internet, i.e., 14 network hops away. We conclude that memory deduplication must also be considered a security risk if only applied within a single security domain.
Cloud computing paradigms have emerged as a major facility to store and process the massive data produced by various business units, public organizations, Internet-of-Things (IoT), and cyber-physical systems (CPS). To meet users' performance requirements while maximizing resource utilization to achieve cost-efficiency, cloud administrators leverage schedulers to orchestrate tasks to different physical nodes and allow applications from different users to share the same physical node. On the other hand, micro-architectural attacks, e.g, side-channel attacks, transient execution attacks, and Rowhammer attacks, exploit the shared resources to compromise the confidentiality/integrity of a co-located victim application. Since co-location is an essential requirement for micro-architectural attacks, in this work, we investigate whether attackers can exploit the cloud schedulers to satisfy the co-location requirement of the micro-architectural attacks. Specifically, in this paper, we comprehensively analyze if attackers can influence the scheduling process of cloud schedulers to co-locate with specific targeted applications in the cloud. Our analysis shows that for cloud schedulers that allow users to submit application requirements, an attacker can carefully select the attacker's application requirements to influence the scheduler to co-locate it with a targeted victim application. We call such attack textit{Rep}lication Atextit{ttack} (Repttack). Our experimental results, in both a simulated cluster environment and a real cluster, show similar trends; a single attack instance can reach up to $50%$ co-location rate (probability of co-location) and with only $5$ instances the co-location rate can reach up to $80%$ in a heterogeneous cloud. Furthermore, we propose and evaluate a mitigation strategy that can help defend against Repttack. We believe that our results highlight the fact that schedulers in multi-user clusters need to be more carefully designed with security in mind, and the process of making scheduling decisions should involve as little user-defined information as possible.
It is well-known that GPS is vulnerable to signal spoofing attacks. Although several spoofing detection techniques exist, they are incapable of mitigation and recovery from stealthy attackers. In this work, we present SemperFi, a single antenna GPS receiver capable of tracking legitimate GPS satellite signals and estimating the true location even against strong adversaries. Our design leverages a combination of the Extended Kalman Filter based GPS failsafe mechanism built into majority of UAVs and a custom designed legitimate signal retriever module to detect and autonomously recover from majority of spoofing attacks. We develop algorithms to carefully synthesize recovery signals and extend the successive interference cancellation technique to preserve the legitimate signal’s ToA, while eliminating the attacker’s signal. For strong adversaries capable of stealthy and seamless takeover, SemperFi uses brief maneuvers designed to exploit the short-term stability of inertial sensors and identify stealthy spoofing attacks. We implement SemperFi in GNSS-SDR, an open-source software-defined GNSS receiver, and evaluate its performance using UAV simulators, real drones, a variety of real-world GPS datasets, as well as on various embedded platforms. Our evaluation results indicate that in many scenarios, SemperFi can identify adversarial peaks by executing flight patterns less than 100 m long and recover the true location within 0.54 s (Jetson Xavier). We show that our receiver is secure against both naive and stealthy spoofers who exploit inertial sensor errors and execute seamless takeover attacks. Furthermore, we design SemperFi as a pluggable module capable of generating a spoofer-free GPS signal for processing on any commercial-off-the-shelf GPS receiver available today. Finally, we release our implementation to the community for usage and further research.
Covert channels are a method of communication that is used to exfiltrate information from computing devices and break the security policy of computer systems. Any shared resource can be potentially leveraged as a covert channel, and conventional wisdom of cyber-security believes that air-gapped computing devices, disconnected from the Internet, are highly secured. Recent studies show that advanced covert channel attacks using acoustic, thermal, and electromagnetic effects can only work under a limited proximity constraint (e.g., within 2 meters). In this work, we present SpiralSpy, a new covert channel to attack air-gapped computing devices through millimeter-wave (mmWave) sensing technologies. SpiralSpy can be stealthily launched and circumvent strongly isolated computing devices from a practical distance (up to 8 meters). Specifically, we demonstrate that ordinal cooling fans can be leveraged for covert channel attacks. A malicious software inside air-gapped computing devices can saliently encode confidential data into the fan control signals, and modulated status on fan motions can be remotely decoded by a commodity mmWave sensor. SpiralSpy can be adopted on multiple-fan systems and enable a scalable capacity for multi-channel and high-speed information transfer. We evaluate SpiralSpy with 71 computing devices with cooling fans. Experimental results demonstrate that SpiralSpy can achieve up to 6 bps that is 6-24X faster than existing covert channels on air-gapped computing devices. We evaluate the usability and robustness of SpiralSpy under different real-world scenarios. Moreover, we conduct in-depth analysis and discussion on countermeasures for SpiralSpy-based covert channel attacks to improve computer and information security.
We analyzed the generation of protocol header fields in the implementations of multiple TCP/IP network stacks and found new ways to leak information about global protocol states. We then demonstrated new covert channels by remotely observing and modifying the system's global state via these protocol fields. Unlike earlier works, our research focuses on hosts that reside in firewalled networks (including source address validation – SAV), which is a very common scenario nowadays. Our attacks are designed to be non-disruptive – in the exfiltration scenario, this makes the attacks stealthier and thus extends their longevity, and in case of host alias resolution and similar techniques – this ensures the techniques are ethical. We focused on ICMP, which is commonly served by firewalls, and on UDP, which is forecasted to take a more prominent share of the Internet traffic with the advent of HTTP/3 and QUIC, though we report results for TCP as well. The information leakage scenarios we discovered enable the construction of practical covert channels which directly pierce firewalls, or indirectly establish communication via hosts in firewalled networks that also employ SAV. We describe and test three novel attacks in this context: exfiltration via the firewall itself, exfiltration via a DMZ host, and exfiltration via co-resident containers. These are three generic, new use cases for covert channels that work around firewalling and enable devices that are not allowed direct communication with the Internet, to still exfiltrate data out of the network. In other words, we exfiltrate data from isolated networks to the Internet. We also explain how to mount known attacks such as host alias resolution, de-NATting and container co-residence detection, using the new information leakage techniques.