Biblio

We present CrypTFlow, a first of its kind system that converts TensorFlow inference code into Secure Multi-party Computation (MPC) protocols at the push of a button. To do this, we build three components. Our first component, Athos, is an end-to-end compiler from TensorFlow to a variety of semihonest MPC protocols. The second component, Porthos, is an improved semi-honest 3-party protocol that provides significant speedups for TensorFlow like applications. Finally, to provide malicious secure MPC protocols, our third component, Aramis, is a novel technique that uses hardware with integrity guarantees to convert any semi-honest MPC protocol into an MPC protocol that provides malicious security. The malicious security of the protocols output by Aramis relies on integrity of the hardware and semi-honest security of MPC. Moreover, our system matches the inference accuracy of plaintext TensorFlow.We experimentally demonstrate the power of our system by showing the secure inference of real-world neural networks such as ResNet50 and DenseNet121 over the ImageNet dataset with running times of about 30 seconds for semi-honest security and under two minutes for malicious security. Prior work in the area of secure inference has been limited to semi-honest security of small networks over tiny datasets such as MNIST or CIFAR. Even on MNIST/CIFAR, CrypTFlow outperforms prior work.

Inferring behavior model of a running software system is quite useful for several automated software engineering tasks, such as program comprehension, anomaly detection, and testing. Most existing dynamic model inference techniques are white-box, i.e., they require source code to be instrumented to get run-time traces. However, in many systems, instrumenting the entire source code is not possible (e.g., when using black-box third-party libraries) or might be very costly. Unfortunately, most black-box techniques that detect states over time are either univariate, or make assumptions on the data distribution, or have limited power for learning over a long period of past behavior. To overcome the above issues, in this paper, we propose a hybrid deep neural network that accepts as input a set of time series, one per input/output signal of the system, and applies a set of convolutional and recurrent layers to learn the non-linear correlations between signals and the patterns, over time. We have applied our approach on a real UAV auto-pilot solution from our industry partner with half a million lines of C code. We ran 888 random recent system-level test cases and inferred states, over time. Our comparison with several traditional time series change point detection techniques showed that our approach improves their performance by up to 102%, in terms of finding state change points, measured by F1 score. We also showed that our state classification algorithm provides on average 90.45% F1 score, which improves traditional classification algorithms by up to 17%.

One of the main threats to the underwater communication cables identified in the recent years is possible tampering or damage by malicious actors. This paper proposes a solution with explanation abilities to detect and investigate this kind of threat within the evidence theory framework. The reasoning scheme implements the traditional “opportunity-capability-intent” threat model to assess a degree to which a given vessel may pose a threat. The scenario discussed considers a variety of possible pieces of information available from different sources. A source quality model is used to reason with the partially reliable sources and the impact of this meta-information on the overall assessment is illustrated. Examples of uncertain relationships between the relevant variables are modelled and the constructed model is used to investigate the probability of threat of four vessels of different types. One of these cases is discussed in more detail to demonstrate the explanation abilities. Explanations about inference are provided thanks to sensitivity spaces in which the impact of the different pieces of information on the reasoning are compared.

This paper proposes a novel extension of blockchain systems with fuzzy ontologies. The main advantage is to let the users have flexible restrictions, represented using fuzzy sets, and to develop smart contracts where there is a partial agreement among the involved parts. We propose a general architecture based on four fuzzy ontologies and a process to develop and run the smart contracts, based on a reduction to a well-known fuzzy ontology reasoning task (Best Satisfiability Degree). We also investigate different operators to compute Pareto-optimal solutions and implement our approach in the Ethereum blockchain.

Trust of image sensor data is becoming increasingly important as the Internet of Things (IoT) applications grow from home appliances to surveillance. Up to our knowledge, there exists only one work in literature that estimates trustworthiness of digital images applied to forensic applications, based on a machine learning technique. The efficacy of this technique is heavily dependent on availability of an appropriate training set and adequate variation of IoT sensor data with noise, interference and environmental condition, but availability of such data cannot be assured always. Therefore, to overcome this limitation, a robust method capable of estimating trustworthy measure with high accuracy is needed. Lowering cost of sensors allow many IoT applications to use multiple types of sensors to observe the same event. In such cases, complementary multimodal data of one sensor can be exploited to measure trust level of another sensor data. In this paper, for the first time, we introduce a completely new approach to estimate the trustworthiness of an image sensor data using another sensor's numerical data. We develop a theoretical model using the Dempster-Shafer theory (DST) framework. The efficacy of the proposed model in estimating trust level of an image sensor data is analyzed by observing a fire event using IoT image and temperature sensor data in a residential setup under different scenarios. The proposed model produces highly accurate trust level in all scenarios with authentic and forged image data.

{Surrogate model-based optimization algorithm remains as an important solution to expensive black-box function optimization. The introduction of ensemble model enables the algorithm to automatically choose a proper model integration mode and adapt to various parameter spaces when dealing with different problems. However, this also significantly increases the computational burden of the algorithm. On the other hand, utilizing parallel computing resources and improving efficiency of black-box function optimization also require combination with surrogate optimization algorithm in order to design and realize an efficient parallel parameter space sampling mechanism. This paper makes use of parallel computing technology to speed up the weight updating related computation for the ensemble model based on Dempster-Shafer theory, and combines it with stochastic response surface method to develop a novel parallel sampling mechanism for asynchronous parameter optimization. Furthermore, it designs and implements corresponding parallel computing framework and applies the developed algorithm to quantitative trading strategy tuning in financial market. It is verified that the algorithm is both feasible and effective in actual application. The experiment demonstrates that with guarantee of optimizing performance, the parallel optimization algorithm can achieve excellent accelerating effect.

In order to standardize and describe vulnerability information in detail as far as possible and realize knowledge sharing, reuse and extension at the semantic level, a vulnerability ontology is constructed based on the information security public databases such as CVE, CWE and CAPEC and industry public standards like CVSS. By analyzing the relationship between vulnerability class and weakness class, inference rules are defined to realize knowledge inference from vulnerability instance to its consequence and from one vulnerability instance to another vulnerability instance. The experimental results show that this model can analyze the causal and congeneric relationships between vulnerability instances, which is helpful to repair vulnerabilities and predict attacks.

Trust mechanisms are considered the logical protection of software systems, preventing malicious people from taking advantage or cheating others. Although these concepts are widely used, most applications in this field do not consider affective aspects to aid in trust computation. Researchers of Psychology, Neurology, Anthropology, and Computer Science argue that affective aspects are essential to human's decision-making processes. So far, there is a lack of understanding about how these aspects impact user's trust, particularly when they are inserted in an evaluation system. In this paper, we propose a trust model that accounts for personality using three personality models: Big Five, Needs, and Values. We tested our approach by extracting personality aspects from texts provided by two online human-fed evaluation systems and correlating them to reputation values. The empirical experiments show statistically significant better results in comparison to non-personality-wise approaches.

Because of the high-value data of an enterprise, sophisticated cyber-attacks targeted at enterprise networks have become prominent. Attack graphs are useful tools that facilitate a scalable security analysis of enterprise networks. However, the administrators face difficulties in effectively modelling security problems and making right decisions when constructing attack graphs as their risk assessment experience is often limited. In this paper, we propose an innovative method of security assessment through an ontology- and graph-based approach. An ontology is designed to represent security knowledge such as assets, vulnerabilities, attacks, countermeasures, and relationships between them in a common vocabulary. An efficient algorithm is proposed to generate an attack graph based on the inference ability of the security ontology. The proposed algorithm is evaluated with different sizes and topologies of test networks; the results show that our proposed algorithm facilitates a scalable security analysis of enterprise networks.

Identifying cyberattack vectors on cyber supply chains (CSC) in the event of cyberattacks are very important in mitigating cybercrimes effectively on Cyber Physical Systems CPS. However, in the cyber security domain, the invincibility nature of cybercrimes makes it difficult and challenging to predict the threat probability and impact of cyber attacks. Although cybercrime phenomenon, risks, and treats contain a lot of unpredictability's, uncertainties and fuzziness, cyberattack detection should be practical, methodical and reasonable to be implemented. We explore Bayesian Belief Networks (BBN) as knowledge representation in artificial intelligence to be able to be formally applied probabilistic inference in the cyber security domain. The aim of this paper is to use Bayesian Belief Networks to detect cyberattacks on CSC in the CPS domain. We model cyberattacks using DAG method to determine the attack propagation. Further, we use a smart grid case study to demonstrate the applicability of attack and the cascading effects. The results show that BBN could be adapted to determine uncertainties in the event of cyberattacks in the CSC domain.

To improve production efficiency, more industrial control systems are connected to IT networks, and more IT technologies are applied to industrial control networks, network security has become an important problem. Industrial control network security analysis and decision-making is a effective method to solve the problem, which can predict risks and support to make decisions before the actual fault of the industrial control network system has not occurred. This paper proposes a security analysis and decision-making method with forward reasoning based on strong relevant logic for industrial control networks. The paper presents a case study in security analysis and decision-making for industrial control networks. The result of the case study shows that the proposed method is effective.

Attack graphs are a relatively new - at least, from the point of view of a practical usage - method for modeling multistage cyber-attacks. They allow to understand how seemingly unrelated vulnerabilities may be combined together by an attacker to form a chain of hostile actions that enable to compromise a key resource. An attack graph is also the starting point for providing recommendations for corrective actions that would fix or mask security problems and prevent the attacks. In the paper, we propose TAG, a topological attack graph analysis tool designed to support a user in a security evaluation and countermeasure selection. TAG employs an improved version of MulVAL inference engine, estimates a security level on the basis of attack graph and attack paths scoring, and recommends remedial actions that improve the security of the analyzed system.

The lack of knowledge about API correct usage rules is one of the main reasons that APIs are employed incorrectly by programmers, which in some cases lead to serious security vulnerabilities. However, finding a correct usage rule for an API is a time-consuming and error-prone task, particularly in the absence of an API documentation. Existing approaches to extract correct usage rules are mostly based on majority API usages, assuming the correct usage is prevalent. Although statistically extracting API correct usage rules achieves reasonable accuracy, it cannot work correctly in the absence of a fair amount of sample usages. We propose inferring API correct usage rules independent of the number of sample usages by leveraging an API tree structure. In an API tree, each node is an API, and each node's children are APIs called by the parent API. Starting from lower-level APIs, it is possible to infer the correct usage rules for them by utilizing the available correct usage rules of their children. We developed a tool based on our idea for inferring API correct usages rules hierarchically, and have applied it to the source code of Linux kernel v4.3 drivers and found 24 previously reported bugs.

Machine Learning as a Service (MLaaS) is becoming a popular practice where Service Consumers, e.g., end-users, send their data to a ML Service and receive the prediction outputs. However, the emerging usage of MLaaS has raised severe privacy concerns about users' proprietary data. PrivacyPreserving Machine Learning (PPML) techniques aim to incorporate cryptographic primitives such as Homomorphic Encryption (HE) and Multi-Party Computation (MPC) into ML services to address privacy concerns from a technology standpoint. Existing PPML solutions have not been widely adopted in practice due to their assumed high overhead and integration difficulty within various ML front-end frameworks as well as hardware backends. In this work, we propose PlaidML-HE, the first end-toend HE compiler for PPML inference. Leveraging the capability of Domain-Specific Languages, PlaidML-HE enables automated generation of HE kernels across diverse types of devices. We evaluate the performance of PlaidML-HE on different ML kernels and demonstrate that PlaidML-HE greatly reduces the overhead of the HE primitive compared to the existing implementations.

While enjoying the convenience brought by mobile phones, users have been exposed to high risk of private information leakage. It is known that many applications on mobile devices read private data and send them to remote servers. However how, when and in what scale the private data are leaked are not investigated systematically in the real-world scenario. In this paper, a framework is proposed to analyze the usage data from mobile devices and the traffic data from the mobile network and make a comprehensive privacy leakage detection and privacy inference mining on a large scale of realworld mobile data. Firstly, this paper sets up a training dataset and trains a privacy detection model on mobile traffic data. Then classical machine learning tools are used to discover private usage patterns. Based on our experiments and data analysis, it is found that i) a large number of private information is transmitted in plaintext, and even passwords are transmitted in plaintext by some applications, ii) more privacy types are leaked in Android than iOS, while GPS location is the most leaked privacy in both Android and iOS system, iii) the usage pattern is related to mobile device price. Through our experiments and analysis, it can be concluded that mobile privacy leakage is pervasive and serious.

In internet based communication, various types of attacks have been evolved. Hence, attacker easily breaches the securities. Traditional intrusion detection techniques to observe these attacks have failed and thus hefty systems are required to remove these attacks before they expose entire network. With the ability of artificial intelligence systems to adapt high computational speed, boost fault tolerance, and error resilience against noisy information, a hybrid particle swarm optimization(PSO) fuzzy rule based inference engine has been designed in this paper. The fuzzy logic based on degree of truth while the PSO algorithm based on population stochastic technique helps in learning from the scenario, thus their combination will increase the toughness of intrusion detection system. The proposed network intrusion detection system will be able to classify normal as well as anomalism behaviour in the network. DARPA-KDD99 dataset examined on this system to address the behaviour of each connection on network and compared with existing system. This approach improves the result on the basis of precision, recall and F1-score.

The smart grid is a complex cyber-physical system (CPS) that poses challenges related to scale, integration, interoperability, processes, governance, and human elements. The US National Institute of Standards and Technology (NIST) and its government, university and industry collaborators, developed an approach, called CPS Framework, to reasoning about CPS across multiple levels of concern and competency, including trustworthiness, privacy, reliability, and regulatory. The approach uses ontology and reasoning techniques to achieve a greater understanding of the interdependencies among the elements of the CPS Framework model applied to use cases. This paper demonstrates that the approach extends naturally to automated and manual decision-making for smart grids: we apply it to smart grid use cases, and illustrate how it can be used to analyze grid topologies and address concerns about the smart grid. Smart grid stakeholders, whose decision making may be assisted by this approach, include planners, designers and operators.

In order to protect individuals' privacy, data have to be "well-sanitized" before sharing them, i.e. one has to remove any personal information before sharing data. However, it is not always clear when data shall be deemed well-sanitized. In this paper, we argue that the evaluation of sanitized data should be based on whether the data allows the inference of sensitive information that is specific to an individual, instead of being centered around the concept of re-identification. We propose a framework to evaluate the effectiveness of different sanitization techniques on a given dataset by measuring how much an individual's record from the sanitized dataset influences the inference of his/her own sensitive attribute. Our intent is not to accurately predict any sensitive attribute but rather to measure the impact of a single record on the inference of sensitive information. We demonstrate our approach by sanitizing two real datasets in different privacy models and evaluate/compare each sanitized dataset in our framework.

Cyber reconnaissance is the process of gathering information about a target network for the purpose of compromising systems within that network. Network-based deception has emerged as a promising approach to disrupt attackers' reconnaissance efforts. However, limited work has been done so far on measuring the effectiveness of network-based deception. Furthermore, given that Software-Defined Networking (SDN) facilitates cyber deception by allowing network traffic to be modified and injected on-the-fly, understanding the effectiveness of employing different cyber deception strategies is critical. In this paper, we present a model to study the reconnaissance surface of a network and model the process of gathering information by attackers as interactions with a cyber defensive system that may use deception. To capture the evolution of the attackers' knowledge during reconnaissance, we design a belief system that is updated by using a Bayesian inference method. For the proposed model, we present two metrics based on KL-divergence to quantify the effectiveness of network deception. We tested the model and the two metrics by conducting experiments with a simulated attacker in an SDN-based deception system. The results of the experiments match our expectations, providing support for the model and proposed metrics.

Inference of unknown opinions with uncertain, adversarial (e.g., incorrect or conflicting) evidence in large datasets is not a trivial task. Without proper handling, it can easily mislead decision making in data mining tasks. In this work, we propose a highly scalable opinion inference probabilistic model, namely Adversarial Collective Opinion Inference (Adv-COI), which provides a solution to infer unknown opinions with high scalability and robustness under the presence of uncertain, adversarial evidence by enhancing Collective Subjective Logic (CSL) which is developed by combining SL and Probabilistic Soft Logic (PSL). The key idea behind the Adv-COI is to learn a model of robust ways against uncertain, adversarial evidence which is formulated as a min-max problem. We validate the out-performance of the Adv-COI compared to baseline models and its competitive counterparts under possible adversarial attacks on the logic-rule based structured data and white and black box adversarial attacks under both clean and perturbed semi-synthetic and real-world datasets in three real world applications. The results show that the Adv-COI generates the lowest mean absolute error in the expected truth probability while producing the lowest running time among all.

Creative works, whether paintings or memes, follow unique journeys that result in their final form. Understanding these journeys, a process known as "provenance analysis," provides rich insights into the use, motivation, and authenticity underlying any given work. The application of this type of study to the expanse of unregulated content on the Internet is what we consider in this paper. Provenance analysis provides a snapshot of the chronology and validity of content as it is uploaded, re-uploaded, and modified over time. Although still in its infancy, automated provenance analysis for online multimedia is already being applied to different types of content. Most current works seek to build provenance graphs based on the shared content between images or videos. This can be a computationally expensive task, especially when considering the vast influx of content that the Internet sees every day. Utilizing non-content-based information, such as timestamps, geotags, and camera IDs can help provide important insights into the path a particular image or video has traveled during its time on the Internet without large computational overhead. This paper tests the scope and applicability of metadata-based inferences for provenance graph construction in two different scenarios: digital image forensics and cultural analytics.

Model compression is considered to be an effective way to reduce the implementation cost of deep neural networks (DNNs) while maintaining the inference accuracy. Many recent studies have developed efficient model compression algorithms and implementations in accelerators on various devices. Protecting integrity of DNN inference against fault attacks is important for diverse deep learning enabled applications. However, there has been little research investigating the fault resilience of DNNs and the impact of model compression on fault tolerance. In this work, we consider faults on different data types and develop a simulation framework for understanding the fault resiliency of compressed DNN models as compared to uncompressed models. We perform our experiments on two common DNNs, LeNet-5 and VGG16, and evaluate their fault resiliency with different types of compression. The results show that binary quantization can effectively increase the fault resilience of DNN models by 10000x for both LeNet5 and VGG16. Finally, we propose software and hardware mitigation techniques to increase the fault resiliency of DNN models.

Deep neural networks are susceptible to various inference attacks as they remember information about their training data. We design white-box inference attacks to perform a comprehensive privacy analysis of deep learning models. We measure the privacy leakage through parameters of fully trained models as well as the parameter updates of models during training. We design inference algorithms for both centralized and federated learning, with respect to passive and active inference attackers, and assuming different adversary prior knowledge. We evaluate our novel white-box membership inference attacks against deep learning algorithms to trace their training data records. We show that a straightforward extension of the known black-box attacks to the white-box setting (through analyzing the outputs of activation functions) is ineffective. We therefore design new algorithms tailored to the white-box setting by exploiting the privacy vulnerabilities of the stochastic gradient descent algorithm, which is the algorithm used to train deep neural networks. We investigate the reasons why deep learning models may leak information about their training data. We then show that even well-generalized models are significantly susceptible to white-box membership inference attacks, by analyzing state-of-the-art pre-trained and publicly available models for the CIFAR dataset. We also show how adversarial participants, in the federated learning setting, can successfully run active membership inference attacks against other participants, even when the global model achieves high prediction accuracies.

The presence of artificial agents in human social networks is growing. From chatbots to robots, human experience in the developed world is moving towards a socio-technical system in which agents can be technological or biological, with increasingly blurred distinctions between. Given that emotion is a key element of human interaction, enabling artificial agents with the ability to reason about affect is a key stepping stone towards a future in which technological agents and humans can work together. This paper presents work on building intelligent computational agents that integrate both emotion and cognition. These agents are grounded in the well-established social-psychological Bayesian Affect Control Theory (BayesAct). The core idea of BayesAct is that humans are motivated in their social interactions by affective alignment: they strive for their social experiences to be coherent at a deep, emotional level with their sense of identity and general world views as constructed through culturally shared symbols. This affective alignment creates cohesive bonds between group members, and is instrumental for collaborations to solidify as relational group commitments. BayesAct agents are motivated in their social interactions by a combination of affective alignment and decision theoretic reasoning, trading the two off as a function of the uncertainty or unpredictability of the situation. This paper provides a high-level view of dual process theories and advances BayesAct as a plausible, computationally tractable model based in social-psychological and sociological theory.

Currently, organisations find it difficult to design a Decision Support System (DSS) that can predict various operational risks, such as financial and quality issues, with operational risks responsible for significant economic losses and damage to an organisation's reputation in the market. This paper proposes a new DSS for risk assessment, called the Fuzzy Inference DSS (FIDSS) mechanism, which uses fuzzy inference methods based on an organisation's big data collection. It includes the Emerging Association Patterns (EAP) technique that identifies the important features of each risk event. Then, the Mamdani fuzzy inference technique and several membership functions are evaluated using the firm's data sources. The FIDSS mechanism can enhance an organisation's decision-making processes by quantifying the severity of a risk as low, medium or high. When it automatically predicts a medium or high level, it assists organisations in taking further actions that reduce this severity level.