Benjamin Zi Hao Zhao

Benjamin Tag, Niels van Berkel, Sunny Verma et al.

Artificial Intelligence (AI) systems have been increasingly used to make decision-making processes faster, more accurate, and more efficient. However, such systems are also at constant risk of being attacked. While the majority of attacks targeting AI-based applications aim to manipulate classifiers or training data and alter the output of an AI model, recently proposed Sponge Attacks against AI models aim to impede the classifier's execution by consuming substantial resources. In this work, we propose \textit{Dual Denial of Decision (DDoD) attacks against collaborative Human-AI teams}. We discuss how such attacks aim to deplete \textit{both computational and human} resources, and significantly impair decision-making capabilities. We describe DDoD on human and computational resources and present potential risk scenarios in a series of exemplary domains.

Conor Atkins, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

One of the new developments in chit-chat bots is a long-term memory mechanism that remembers information from past conversations for increasing engagement and consistency of responses. The bot is designed to extract knowledge of personal nature from their conversation partner, e.g., stating preference for a particular color. In this paper, we show that this memory mechanism can result in unintended behavior. In particular, we found that one can combine a personal statement with an informative statement that would lead the bot to remember the informative statement alongside personal knowledge in its long term memory. This means that the bot can be tricked into remembering misinformation which it would regurgitate as statements of fact when recalling information relevant to the topic of conversation. We demonstrate this vulnerability on the BlenderBot 2 framework implemented on the ParlAI platform and provide examples on the more recent and significantly larger BlenderBot 3 model. We generate 150 examples of misinformation, of which 114 (76%) were remembered by BlenderBot 2 when combined with a personal statement. We further assessed the risk of this misinformation being recalled after intervening innocuous conversation and in response to multiple questions relevant to the injected memory. Our evaluation was performed on both the memory-only and the combination of memory and internet search modes of BlenderBot 2. From the combinations of these variables, we generated 12,890 conversations and analyzed recalled misinformation in the responses. We found that when the chat bot is questioned on the misinformation topic, it was 328% more likely to respond with the misinformation as fact when the misinformation was in the long-term memory.

Muhammad Salman, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

Adversarial examples add imperceptible alterations to inputs with the objective to induce misclassification in machine learning models. They have been demonstrated to pose significant challenges in domains like image classification, with results showing that an adversarially perturbed image to evade detection against one classifier is most likely transferable to other classifiers. Adversarial examples have also been studied in malware analysis. Unlike images, program binaries cannot be arbitrarily perturbed without rendering them non-functional. Due to the difficulty of crafting adversarial program binaries, there is no consensus on the transferability of adversarially perturbed programs to different detectors. In this work, we explore the robustness of malware detectors against adversarially perturbed malware. We investigate the transferability of adversarial attacks developed against one detector, against other machine learning-based malware detectors, and code similarity techniques, specifically, locality sensitive hashing-based detectors. Our analysis reveals that adversarial program binaries crafted for one detector are generally less effective against others. We also evaluate an ensemble of detectors and show that they can potentially mitigate the impact of adversarial program binaries. Finally, we demonstrate that substantial program changes made to evade detection may result in the transformation technique being identified, implying that the adversary must make minimal changes to the program binary.

Rana Salal Ali, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

Named entity recognition models (NER), are widely used for identifying named entities (e.g., individuals, locations, and other information) in text documents. Machine learning based NER models are increasingly being applied in privacy-sensitive applications that need automatic and scalable identification of sensitive information to redact text for data sharing. In this paper, we study the setting when NER models are available as a black-box service for identifying sensitive information in user documents and show that these models are vulnerable to membership inference on their training datasets. With updated pre-trained NER models from spaCy, we demonstrate two distinct membership attacks on these models. Our first attack capitalizes on unintended memorization in the NER's underlying neural network, a phenomenon NNs are known to be vulnerable to. Our second attack leverages a timing side-channel to target NER models that maintain vocabularies constructed from the training data. We show that different functional paths of words within the training dataset in contrast to words not previously seen have measurable differences in execution time. Revealing membership status of training samples has clear privacy implications, e.g., in text redaction, sensitive words or phrases to be found and removed, are at risk of being detected in the training dataset. Our experimental evaluation includes the redaction of both password and health data, presenting both security risks and privacy/regulatory issues. This is exacerbated by results that show memorization with only a single phrase. We achieved 70% AUC in our first attack on a text redaction use-case. We also show overwhelming success in the timing attack with 99.23% AUC. Finally we discuss potential mitigation approaches to realize the safe use of NER models in light of the privacy and security implications of membership inference attacks.

Shashie Dilhara Batan Arachchige, Hassan Jameel Asghar, Benjamin Zi Hao Zhao et al.

Large Language Models (LLMs) generate responses based on user prompts. Often, these prompts may contain highly sensitive information, including personally identifiable information (PII), which could be exposed to third parties hosting these models. In this work, we propose a new method to sanitize user prompts. Our mechanism uses the randomized response mechanism of differential privacy to randomly and independently perturb each character in a word. The perturbed text is then sent to a remote LLM, which first performs a prompt restoration and subsequently performs the intended downstream task. The idea is that the restoration will be able to reconstruct non-sensitive words even when they are perturbed due to cues from the context, as well as the fact that these words are often very common. On the other hand, perturbation would make reconstruction of sensitive words difficult because they are rare. We experimentally validate our method on two datasets, i2b2/UTHealth and Enron, using two LLMs: Llama-3.1 8B Instruct and GPT-4o mini. We also compare our approach with a word-level differentially private mechanism, and with a rule-based PII redaction baseline, using a unified privacy-utility evaluation. Our results show that sensitive PII tagged in these datasets are reconstructed at a rate close to the theoretical rate of reconstructing completely random words, whereas non-sensitive words are reconstructed at a much higher rate. Our method has the advantage that it can be applied without explicitly identifying sensitive pieces of information in the prompt, while showing a good privacy-utility tradeoff for downstream tasks.

Shashie Dilhara Batan Arachchige, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

Large Language Models (LLMs) are often fine-tuned to adapt their general-purpose knowledge to specific tasks and domains such as cyber threat intelligence (CTI). Fine-tuning is mostly done through proprietary datasets that may contain sensitive information. Owners expect their fine-tuned model to not inadvertently leak this information to potentially adversarial end users. Using CTI as a use case, we demonstrate that data-extraction attacks can recover sensitive information from fine-tuned models on CTI reports, underscoring the need for mitigation. Retraining the full model to eliminate this leakage is computationally expensive and impractical. We propose an alternative approach, which we call privacy alignment, inspired by safety alignment in LLMs. Just like safety alignment teaches the model to abide by safety constraints through a few examples, we enforce privacy alignment through few-shot supervision, integrating a privacy classifier and a privacy redactor, both handled by the same underlying LLM. We evaluate our system, called CTIGuardian, using GPT-4o mini and Mistral-7B Instruct models, benchmarking against Presidio, a named entity recognition (NER) baseline. Results show that CTIGuardian provides a better privacy-utility trade-off than NER based models. While we demonstrate its effectiveness on a CTI use case, the framework is generic enough to be applicable to other sensitive domains.

Ali Reza Ghavamipour, Benjamin Zi Hao Zhao, Oguzhan Ersoy et al.

Decentralized machine learning (DL) has been receiving an increasing interest recently due to the elimination of a single point of failure, present in Federated learning setting. Yet, it is threatened by the looming threat of Byzantine clients who intentionally disrupt the learning process by broadcasting arbitrary model updates to other clients, seeking to degrade the performance of the global model. In response, robust aggregation schemes have emerged as promising solutions to defend against such Byzantine clients, thereby enhancing the robustness of Decentralized Learning. Defenses against Byzantine adversaries, however, typically require access to the updates of other clients, a counterproductive privacy trade-off that in turn increases the risk of inference attacks on those same model updates. In this paper, we introduce SecureDL, a novel DL protocol designed to enhance the security and privacy of DL against Byzantine threats. SecureDL~facilitates a collaborative defense, while protecting the privacy of clients' model updates through secure multiparty computation. The protocol employs efficient computation of cosine similarity and normalization of updates to robustly detect and exclude model updates detrimental to model convergence. By using MNIST, Fashion-MNIST, SVHN and CIFAR-10 datasets, we evaluated SecureDL against various Byzantine attacks and compared its effectiveness with four existing defense mechanisms. Our experiments show that SecureDL is effective even in the case of attacks by the malicious majority (e.g., 80% Byzantine clients) while preserving high training accuracy.

Ali Reza Ghavamipour, Benjamin Zi Hao Zhao, Fatih Turkmen

Decentralized learning (DL) offers a novel paradigm in machine learning by distributing training across clients without central aggregation, enhancing scalability and efficiency. However, DL's peer-to-peer model raises challenges in protecting against inference attacks and privacy leaks. By forgoing central bottlenecks, DL demands privacy-preserving aggregation methods to protect data from 'honest but curious' clients and adversaries, maintaining network-wide privacy. Privacy-preserving DL faces the additional hurdle of client dropout, clients not submitting updates due to connectivity problems or unavailability, further complicating aggregation. This work proposes three secret sharing-based dropout resilience approaches for privacy-preserving DL. Our study evaluates the efficiency, performance, and accuracy of these protocols through experiments on datasets such as MNIST, Fashion-MNIST, SVHN, and CIFAR-10. We compare our protocols with traditional secret-sharing solutions across scenarios, including those with up to 1000 clients. Evaluations show that our protocols significantly outperform conventional methods, especially in scenarios with up to 30% of clients dropout and model sizes of up to $10^6$ parameters. Our approaches demonstrate markedly high efficiency with larger models, higher dropout rates, and extensive client networks, highlighting their effectiveness in enhancing decentralized learning systems' privacy and dropout robustness.

Robin Carpentier, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

Interactions with online Large Language Models raise privacy issues where providers can gather sensitive information about users and their companies from the prompts. While textual prompts can be sanitized using Differential Privacy, we show that it is difficult to anticipate the performance of an LLM on such sanitized prompt. Poor performance has clear monetary consequences for LLM services charging on a pay-per-use model as well as great amount of computing resources wasted. To this end, we propose a middleware architecture leveraging a Small Language Model to predict the utility of a given sanitized prompt before it is sent to the LLM. We experimented on a summarization task and a translation task to show that our architecture helps prevent such resource waste for up to 20% of the prompts. During our study, we also reproduced experiments from one of the most cited paper on text sanitization using DP and show that a potential performance-driven implementation choice dramatically changes the output while not being explicitly acknowledged in the paper.

Sunday Oyinlola Ogundoyin, Muhammad Ikram, Hassan Jameel Asghar et al.

Millions of users leverage generative pretrained transformer (GPT)-based language models developed by leading model providers for a wide range of tasks. To support enhanced user interaction and customization, many platforms-such as OpenAI-now enable developers to create and publish tailored model instances, known as custom GPTs, via dedicated repositories or application stores. These custom GPTs empower users to browse and interact with specialized applications designed to meet specific needs. However, as custom GPTs see growing adoption, concerns regarding their security vulnerabilities have intensified. Existing research on these vulnerabilities remains largely theoretical, often lacking empirical, large-scale, and statistically rigorous assessments of associated risks. In this study, we analyze 14,904 custom GPTs to assess their susceptibility to seven exploitable threats, such as roleplay-based attacks, system prompt leakage, phishing content generation, and malicious code synthesis, across various categories and popularity tiers within the OpenAI marketplace. We introduce a multi-metric ranking system to examine the relationship between a custom GPT's popularity and its associated security risks. Our findings reveal that over 95% of custom GPTs lack adequate security protections. The most prevalent vulnerabilities include roleplay-based vulnerabilities (96.51%), system prompt leakage (92.20%), and phishing (91.22%). Furthermore, we demonstrate that OpenAI's foundational models exhibit inherent security weaknesses, which are often inherited or amplified in custom GPTs. These results highlight the urgent need for enhanced security measures and stricter content moderation to ensure the safe deployment of GPT-based applications.

Chenhan Zhang, Benjamin Zi Hao Zhao, Hassan Asghar et al.

Users may inadvertently upload personally identifiable information (PII) to Machine Learning as a Service (MLaaS) providers. When users no longer want their PII on these services, regulations like GDPR and COPPA mandate a right to forget for these users. As such, these services seek efficient methods to remove the influence of specific data points. Thus the introduction of machine unlearning. Traditionally, unlearning is performed with the removal of entire data samples (sample unlearning) or whole features across the dataset (feature unlearning). However, these approaches are not equipped to handle the more granular and challenging task of unlearning specific objects within a sample. To address this gap, we propose a scene graph-based object unlearning framework. This framework utilizes scene graphs, rich in semantic representation, transparently translate unlearning requests into actionable steps. The result, is the preservation of the overall semantic integrity of the generated image, bar the unlearned object. Further, we manage high computational overheads with influence functions to approximate the unlearning process. For validation, we evaluate the unlearned object's fidelity in outputs under the tasks of image reconstruction and image synthesis. Our proposed framework demonstrates improved object unlearning outcomes, with the preservation of unrequested samples in contrast to sample and feature learning methods. This work addresses critical privacy issues by increasing the granularity of targeted machine unlearning through forgetting specific object-level details without sacrificing the utility of the whole data sample or dataset feature.

Wanchuang Zhu, Benjamin Zi Hao Zhao, Simon Luo et al.

Byzantine attacks hinder the deployment of federated learning algorithms. Although we know that the benign gradients and Byzantine attacked gradients are distributed differently, to detect the malicious gradients is challenging due to (1) the gradient is high-dimensional and each dimension has its unique distribution and (2) the benign gradients and the attacked gradients are always mixed (two-sample test methods cannot apply directly). To address the above, for the first time, we propose MANDERA which is theoretically guaranteed to efficiently detect all malicious gradients under Byzantine attacks with no prior knowledge or history about the number of attacked nodes. More specifically, we transfer the original updating gradient space into a ranking matrix. By such an operation, the scales of different dimensions of the gradients in the ranking space become identical. The high-dimensional benign gradients and the malicious gradients can be easily separated. The effectiveness of MANDERA is further confirmed by experimentation on four Byzantine attack implementations (Gaussian, Zero Gradient, Sign Flipping, Shifted Mean), comparing with state-of-the-art defenses. The experiments cover both IID and Non-IID datasets.

Shaofeng Li, Hui Liu, Tian Dong et al.

Natural language processing (NLP) systems have been proven to be vulnerable to backdoor attacks, whereby hidden features (backdoors) are trained into a language model and may only be activated by specific inputs (called triggers), to trick the model into producing unexpected behaviors. In this paper, we create covert and natural triggers for textual backdoor attacks, \textit{hidden backdoors}, where triggers can fool both modern language models and human inspection. We deploy our hidden backdoors through two state-of-the-art trigger embedding methods. The first approach via homograph replacement, embeds the trigger into deep neural networks through the visual spoofing of lookalike character replacement. The second approach uses subtle differences between text generated by language models and real natural text to produce trigger sentences with correct grammar and high fluency. We demonstrate that the proposed hidden backdoors can be effective across three downstream security-critical NLP tasks, representative of modern human-centric NLP systems, including toxic comment detection, neural machine translation (NMT), and question answering (QA). Our two hidden backdoor attacks can achieve an Attack Success Rate (ASR) of at least $97\%$ with an injection rate of only $3\%$ in toxic comment detection, $95.1\%$ ASR in NMT with less than $0.5\%$ injected data, and finally $91.12\%$ ASR against QA updated with only 27 poisoning data samples on a model previously trained with 92,024 samples (0.029\%). We are able to demonstrate the adversary's high success rate of attacks, while maintaining functionality for regular users, with triggers inconspicuous by the human administrators.

Benjamin Zi Hao Zhao, Aviral Agrawal, Catisha Coburn et al.

With an increase in low-cost machine learning APIs, advanced machine learning models may be trained on private datasets and monetized by providing them as a service. However, privacy researchers have demonstrated that these models may leak information about records in the training dataset via membership inference attacks. In this paper, we take a closer look at another inference attack reported in literature, called attribute inference, whereby an attacker tries to infer missing attributes of a partially known record used in the training dataset by accessing the machine learning model as an API. We show that even if a classification model succumbs to membership inference attacks, it is unlikely to be susceptible to attribute inference attacks. We demonstrate that this is because membership inference attacks fail to distinguish a member from a nearby non-member. We call the ability of an attacker to distinguish the two (similar) vectors as strong membership inference. We show that membership inference attacks cannot infer membership in this strong setting, and hence inferring attributes is infeasible. However, under a relaxed notion of attribute inference, called approximate attribute inference, we show that it is possible to infer attributes close to the true attributes. We verify our results on three publicly available datasets, five membership, and three attribute inference attacks reported in literature.

Liuqiao Chen, Hu Wang, Benjamin Zi Hao Zhao et al.

Deep Neural Networks have achieved unprecedented success in the field of face recognition such that any individual can crawl the data of others from the Internet without their explicit permission for the purpose of training high-precision face recognition models, creating a serious violation of privacy. Recently, a well-known system named Fawkes (published in USENIX Security 2020) claimed this privacy threat can be neutralized by uploading cloaked user images instead of their original images. In this paper, we present Oriole, a system that combines the advantages of data poisoning attacks and evasion attacks, to thwart the protection offered by Fawkes, by training the attacker face recognition model with multi-cloaked images generated by Oriole. Consequently, the face recognition accuracy of the attack model is maintained and the weaknesses of Fawkes are revealed. Experimental results show that our proposed Oriole system is able to effectively interfere with the performance of the Fawkes system to achieve promising attacking results. Our ablation study highlights multiple principal factors that affect the performance of the Oriole system, including the DSSIM perturbation budget, the ratio of leaked clean user images, and the numbers of multi-cloaks for each uncloaked image. We also identify and discuss at length the vulnerabilities of Fawkes. We hope that the new methodology presented in this paper will inform the security community of a need to design more robust privacy-preserving deep learning models.

Benjamin Zi Hao Zhao, Mohamed Ali Kaafar, Nicolas Kourtellis

Data holders are increasingly seeking to protect their user's privacy, whilst still maximizing their ability to produce machine models with high quality predictions. In this work, we empirically evaluate various implementations of differential privacy (DP), and measure their ability to fend off real-world privacy attacks, in addition to measuring their core goal of providing accurate classifications. We establish an evaluation framework to ensure each of these implementations are fairly evaluated. Our selection of DP implementations add DP noise at different positions within the framework, either at the point of data collection/release, during updates while training of the model, or after training by perturbing learned model parameters. We evaluate each implementation across a range of privacy budgets, and datasets, each implementation providing the same mathematical privacy guarantees. By measuring the models' resistance to real world attacks of membership and attribute inference, and their classification accuracy. we determine which implementations provide the most desirable tradeoff between privacy and utility. We found that the number of classes of a given dataset is unlikely to influence where the privacy and utility tradeoff occurs. Additionally, in the scenario that high privacy constraints are required, perturbing input training data does not trade off as much utility, as compared to noise added later in the ML process.

Benjamin Zi Hao Zhao, Hassan Jameel Asghar, Mohamed Ali Kaafar et al.

Observation Resilient Authentication Schemes (ORAS) are a class of shared secret challenge-response identification schemes where a user mentally computes the response via a cognitive function to authenticate herself such that eavesdroppers cannot readily extract the secret. Security evaluation of ORAS generally involves quantifying information leaked via observed challenge-response pairs. However, little work has evaluated information leaked via human behavior while interacting with these schemes. A common way to achieve observation resilience is by including a modulus operation in the cognitive function. This minimizes the information leaked about the secret due to the many-to-one map from the set of possible secrets to a given response. In this work, we show that user behavior can be used as a side-channel to obtain the secret in such ORAS. Specifically, the user's eye-movement patterns and associated timing information can deduce whether a modulus operation was performed (a fundamental design element), to leak information about the secret. We further show that the secret can still be retrieved if the deduction is erroneous, a more likely case in practice. We treat the vulnerability analytically, and propose a generic attack algorithm that iteratively obtains the secret despite the "faulty" modulus information. We demonstrate the attack on five ORAS, and show that the secret can be retrieved with considerably less challenge-response pairs than non-side-channel attacks (e.g., algebraic/statistical attacks). In particular, our attack is applicable on Mod10, a one-time-pad based scheme, for which no non-side-channel attack exists. We field test our attack with a small-scale eye-tracking user study.

Shaofeng Li, Shiqing Ma, Minhui Xue et al.

Intuitively, a backdoor attack against Deep Neural Networks (DNNs) is to inject hidden malicious behaviors into DNNs such that the backdoor model behaves legitimately for benign inputs, yet invokes a predefined malicious behavior when its input contains a malicious trigger. The trigger can take a plethora of forms, including a special object present in the image (e.g., a yellow pad), a shape filled with custom textures (e.g., logos with particular colors) or even image-wide stylizations with special filters (e.g., images altered by Nashville or Gotham filters). These filters can be applied to the original image by replacing or perturbing a set of image pixels.

Jialin Wen, Benjamin Zi Hao Zhao, Minhui Xue et al.

With the rise of third parties in the machine learning pipeline, the service provider in "Machine Learning as a Service" (MLaaS), or external data contributors in online learning, or the retraining of existing models, the need to ensure the security of the resulting machine learning models has become an increasingly important topic. The security community has demonstrated that without transparency of the data and the resulting model, there exist many potential security risks, with new risks constantly being discovered. In this paper, we focus on one of these security risks -- poisoning attacks. Specifically, we analyze how attackers may interfere with the results of regression learning by poisoning the training datasets. To this end, we analyze and develop a new poisoning attack algorithm. Our attack, termed Nopt, in contrast with previous poisoning attack algorithms, can produce larger errors with the same proportion of poisoning data-points. Furthermore, we also significantly improve the state-of-the-art defense algorithm, termed TRIM, proposed by Jagielsk et al. (IEEE S&P 2018), by incorporating the concept of probability estimation of clean data-points into the algorithm. Our new defense algorithm, termed Proda, demonstrates an increased effectiveness in reducing errors arising from the poisoning dataset through optimizing ensemble models. We highlight that the time complexity of TRIM had not been estimated; however, we deduce from their work that TRIM can take exponential time complexity in the worst-case scenario, in excess of Proda's logarithmic time. The performance of both our proposed attack and defense algorithms is extensively evaluated on four real-world datasets of housing prices, loans, health care, and bike sharing services. We hope that our work will inspire future research to develop more robust learning algorithms immune to poisoning attacks.

Benjamin Zi Hao Zhao, Hassan Jameel Asghar, Mohamed Ali Kaafar

We assess the security of machine learning based biometric authentication systems against an attacker who submits uniform random inputs, either as feature vectors or raw inputs, in order to find an accepting sample of a target user. The average false positive rate (FPR) of the system, i.e., the rate at which an impostor is incorrectly accepted as the legitimate user, may be interpreted as a measure of the success probability of such an attack. However, we show that the success rate is often higher than the FPR. In particular, for one reconstructed biometric system with an average FPR of 0.03, the success rate was as high as 0.78. This has implications for the security of the system, as an attacker with only the knowledge of the length of the feature space can impersonate the user with less than 2 attempts on average. We provide detailed analysis of why the attack is successful, and validate our results using four different biometric modalities and four different machine learning classifiers. Finally, we propose mitigation techniques that render such attacks ineffective, with little to no effect on the accuracy of the system.

Shaofeng Li, Minhui Xue, Benjamin Zi Hao Zhao et al.

Deep neural networks (DNNs) have been proven vulnerable to backdoor attacks, where hidden features (patterns) trained to a normal model, which is only activated by some specific input (called triggers), trick the model into producing unexpected behavior. In this paper, we create covert and scattered triggers for backdoor attacks, invisible backdoors, where triggers can fool both DNN models and human inspection. We apply our invisible backdoors through two state-of-the-art methods of embedding triggers for backdoor attacks. The first approach on Badnets embeds the trigger into DNNs through steganography. The second approach of a trojan attack uses two types of additional regularization terms to generate the triggers with irregular shape and size. We use the Attack Success Rate and Functionality to measure the performance of our attacks. We introduce two novel definitions of invisibility for human perception; one is conceptualized by the Perceptual Adversarial Similarity Score (PASS) and the other is Learned Perceptual Image Patch Similarity (LPIPS). We show that the proposed invisible backdoors can be fairly effective across various DNN models as well as four datasets MNIST, CIFAR-10, CIFAR-100, and GTSRB, by measuring their attack success rates for the adversary, functionality for the normal users, and invisibility scores for the administrators. We finally argue that the proposed invisible backdoor attacks can effectively thwart the state-of-the-art trojan backdoor detection approaches, such as Neural Cleanse and TABOR.

Benjamin Zi Hao Zhao, Hassan Jameel Asghar, Raghav Bhaskar et al.

A number of recent works have demonstrated that API access to machine learning models leaks information about the dataset records used to train the models. Further, the work of \cite{somesh-overfit} shows that such membership inference attacks (MIAs) may be sufficient to construct a stronger breed of attribute inference attacks (AIAs), which given a partial view of a record can guess the missing attributes. In this work, we show (to the contrary) that MIA may not be sufficient to build a successful AIA. This is because the latter requires the ability to distinguish between similar records (differing only in a few attributes), and, as we demonstrate, the current breed of MIA are unsuccessful in distinguishing member records from similar non-member records. We thus propose a relaxed notion of AIA, whose goal is to only approximately guess the missing attributes and argue that such an attack is more likely to be successful, if MIA is to be used as a subroutine for inferring training record attributes.

Yuantian Miao, Minhui Xue, Chao Chen et al.

With the rapid development of deep learning techniques, the popularity of voice services implemented on various Internet of Things (IoT) devices is ever increasing. In this paper, we examine user-level membership inference in the problem space of voice services, by designing an audio auditor to verify whether a specific user had unwillingly contributed audio used to train an automatic speech recognition (ASR) model under strict black-box access. With user representation of the input audio data and their corresponding translated text, our trained auditor is effective in user-level audit. We also observe that the auditor trained on specific data can be generalized well regardless of the ASR model architecture. We validate the auditor on ASR models trained with LSTM, RNNs, and GRU algorithms on two state-of-the-art pipelines, the hybrid ASR system and the end-to-end ASR system. Finally, we conduct a real-world trial of our auditor on iPhone Siri, achieving an overall accuracy exceeding 80\%. We hope the methodology developed in this paper and findings can inform privacy advocates to overhaul IoT privacy.

Benjamin Zi Hao Zhao, Muhammad Ikram, Hassan Jameel Asghar et al.

This paper focuses on reporting of Internet malicious activity (or mal-activity in short) by public blacklists with the objective of providing a systematic characterization of what has been reported over the years, and more importantly, the evolution of reported activities. Using an initial seed of 22 blacklists, covering the period from January 2007 to June 2017, we collect more than 51 million mal-activity reports involving 662K unique IP addresses worldwide. Leveraging the Wayback Machine, antivirus (AV) tool reports and several additional public datasets (e.g., BGP Route Views and Internet registries) we enrich the data with historical meta-information including geo-locations (countries), autonomous system (AS) numbers and types of mal-activity. Furthermore, we use the initially labelled dataset of approx 1.57 million mal-activities (obtained from public blacklists) to train a machine learning classifier to classify the remaining unlabeled dataset of approx 44 million mal-activities obtained through additional sources. We make our unique collected dataset (and scripts used) publicly available for further research. The main contributions of the paper are a novel means of report collection, with a machine learning approach to classify reported activities, characterization of the dataset and, most importantly, temporal analysis of mal-activity reporting behavior. Inspired by P2P behavior modeling, our analysis shows that some classes of mal-activities (e.g., phishing) and a small number of mal-activity sources are persistent, suggesting that either blacklist-based prevention systems are ineffective or have unreasonably long update periods. Our analysis also indicates that resources can be better utilized by focusing on heavy mal-activity contributors, which constitute the bulk of mal-activities.

Jagmohan Chauhan, Benjamin Zi Hao Zhao, Hassan Jameel Asghar et al.

We propose that by integrating behavioural biometric gestures---such as drawing figures on a touch screen---with challenge-response based cognitive authentication schemes, we can benefit from the properties of both. On the one hand, we can improve the usability of existing cognitive schemes by significantly reducing the number of challenge-response rounds by (partially) relying on the hardness of mimicking carefully designed behavioural biometric gestures. On the other hand, the observation resistant property of cognitive schemes provides an extra layer of protection for behavioural biometrics; an attacker is unsure if a failed impersonation is due to a biometric failure or a wrong response to the challenge. We design and develop an instantiation of such a "hybrid" scheme, and call it BehavioCog. To provide security close to a 4-digit PIN---one in 10,000 chance to impersonate---we only need two challenge-response rounds, which can be completed in less than 38 seconds on average (as estimated in our user study), with the advantage that unlike PINs or passwords, the scheme is secure under observation.