Binary Neural Networks (BNNs) enable efficient deep learning by saving on storage and computational costs. However, as the size of neural networks continues to grow, meeting computational require- ments remains a challenge. In this work, we propose a new form of quantization to tile neural network layers with sequences of bits to achieve sub-bit compression of binary-weighted neural networks. The method learns binary vectors (i.e. tiles) to populate each layer of a model via aggregation and reshaping operations. During in- ference, the method reuses a single tile per layer to represent the full tensor. We employ the approach to both fully-connected and convolutional layers, which make up the breadth of space in most neural architectures. Empirically, the approach achieves near full- precision performance on a diverse range of architectures (CNNs, Transformers, MLPs) and tasks (classification, segmentation, and time series forecasting) with up to an 8x reduction in size compared to binary-weighted models. We provide two implementations for Tiled Bit Networks: 1) we deploy the model to a microcontroller to assess its feasibility in resource-constrained environments, and 2) a GPU-compatible inference kernel to facilitate the reuse of a single tile per layer in memory.

Learning from the collective knowledge of data dispersed across private sources can provide neural networks with enhanced generalization capabilities. Federated learning, a method for collaboratively training a machine learning model across remote clients, achieves this by combining client models via the orchestration of a central server. However, current approaches face two critical limitations: i) they struggle to converge when client domains are sufficiently different, and ii) current aggregation techniques produce an identical global model for each client. In this work, we address these issues by reformulating the typical federated learning setup: rather than learning a single global model, we learn N models each optimized for a common objective. To achieve this, we apply a weighted distance minimization to model parameters shared in a peer-to-peer topology. The resulting framework, Iterative Parameter Alignment, applies naturally to the cross-silo setting, and has the following properties: (i) a unique solution for each participant, with the option to globally converge each model in the federation, and (ii) an optional early-stopping mechanism to elicit fairness among peers in collaborative learning settings. These characteristics jointly provide a flexible new framework for iteratively learning from peer models trained on disparate datasets. We find that the technique achieves competitive results on a variety of data partitions compared to state-of-the-art approaches. Further, we show that the method is robust to divergent domains (i.e. disjoint classes across peers) where existing approaches struggle.

In this research, the prevailing issue we address is the over-generalized perspective of spam/ham (non-spam) classification. Despite the intricacies of spam classification, reliance on user feedback may inadvertently skew filters to misclassify legitimate and malicious emails, as users are prone to flag innocuous commercial mail as spam rather than unsubscribing. Current spam datasets have a propensity to include such user-flagged spam which can lead to further misclassification, leading to filters biased against warranted commercial correspondence. Motivated to address this concern, we introduce two new classification categories that delve deeper into the nuances of spam. 'Warranted spam', refers to consensual communications, from a credible source with transparent and safe opt-out mechanisms, and 'unwarranted spam' describes unsolicited messages, often of a malicious nature. Utilizing these classifications, we propose an innovative and dynamic 'warranted spam' dataset that seeks to pave the way for researchers to develop more sophisticated spam filtering techniques. Furthermore, our study delves into pioneering machine learning and natural language processing approaches, harnessing our dataset's potential. The overarching aspiration of our work is to augment online safety, preserve brand integrity, and optimize both the user experience and the efficacy of email marketing campaigns.

Compressed Neural Networks have the potential to enable deep learning across new applications and smaller computational environments. However, understanding the range of learning tasks in which such models can succeed is not well studied. In this work, we apply sparse and binary-weighted Transformers to multivariate time series problems, showing that the lightweight models achieve accuracy comparable to that of dense floating-point Transformers of the same structure. Our model achieves favorable results across three time series learning tasks: classification, anomaly detection, and single-step forecasting. Additionally, to reduce the computational complexity of the attention mechanism, we apply two modifications, which show little to no decline in model performance: 1) in the classification task, we apply a fixed mask to the query, key, and value activations, and 2) for forecasting and anomaly detection, which rely on predicting outputs at a single point in time, we propose an attention mask to allow computation only at the current time step. Together, each compression technique and attention modification substantially reduces the number of non-zero operations necessary in the Transformer. We measure the computational savings of our approach over a range of metrics including parameter count, bit size, and floating point operation (FLOPs) count, showing up to a 53× reduction in storage size and up to 10.5× reduction in FLOPs.

Access control policies (ACPs) are natural language statements that describe criteria under which users can access resources. We focus on constructing NIST Next Generation Access Control (NGAC) ABAC model from ACP statements. NGAC is more complex than RBAC or XACML ABAC as it supports dynamic, event-based policies, as well as prohibitions. We provide algorithms that use spaCy, a NLP library, to extract entities and relations from ACP sentences and convert them into the NGAC model. We then convert this NGAC model into Neo4j representation for the purpose of analysis. We apply the approach to various real-world ACP datasets to demonstrate the feasibility and assess scalability. We demonstrate that the approach is scalable and effectively extracts the NGAC ABAC model from large ACP datasets. We also show that redundancies and inconsistencies of ACP sentences are often found in unclean datasets.

Modern network infrastructures are in a constant state of transformation, in large part due to the exponential growth of Internet of Things (IoT) devices. The unique properties of IoT-connected networks, such as heterogeneity and non-standardized protocol, have created critical security holes and network mismanagement. In this paper we propose a new measurement tool, Intrinsic Dimensionality (ID), to aid in analyzing and classifying network traffic. A proxy for dataset complexity, ID can be used to understand the network as a whole, aiding in tasks such as network management and provisioning. We use ID to evaluate several modern network datasets empirically. Showing that, for network and device-level data, generated using IoT methodologies, the ID of the data fits into a low dimensional representation. Additionally we explore network data complexity at the sample level using Local Intrinsic Dimensionality (LID) and propose a novel unsupervised intrusion detection technique, the Weighted Hamming LID Estimator. We show that the algortihm performs better on IoT network datasets than the Autoencoder, KNN, and Isolation Forests. Finally, we propose the use of synthetic data as an additional tool for both network data measurement as well as intrusion detection. Synthetically generated data can aid in building a more robust network dataset, while also helping in downstream tasks such as machine learning based intrusion detection models. We explore the effects of synthetic data on ID measurements, as well as its role in intrusion detection systems.

Supervised machine learning is often used to detect phishing websites. However, the scarcity of phishing data for training purposes limits the classifier's performance. Further, machine learning algorithms are prone to adversarial attacks: small perturbations on attack data can bypass the classifier. These problems make machine learning less effective for phishing detection. We propose two Generative Adversarial Network (GAN) based approaches that synthesize phishing and legitimate samples to mimic real-world websites. Information about real-world datasets is obtained from ten publicly available phishing datasets which are used by the AAE (Adversarial Autoencoder) and WGAN (Wasserstein GAN) for generating synthetic data. Using both real and synthesized data, we demonstrate how to implement classifiers with higher performance and more resistance to adversarial attacks. We propose a set of hypotheses and validate them through experiments to demonstrate: (i) indistinguishability of synthesized samples from actual ones, (ii) susceptibility of classifiers to adversarial attacks, (iii) mitigating adversarial attacks by training on larger datasets that include correctly labeled synthesized samples, and (iv) better performance of classifiers trained on large datasets. Our AAE and WGAN have been trained on a wide range of datasets, making us optimistic about its widespread applicability.

Almost all aspects of modern automobiles are controlled by embedded computers, known as Electronic Control Units (ECUs). ECUs are connected with each other over a Controller Area Network (CAN) network. ECUs communicate with each other and control the automobile's behavior using messages. Heavy vehicles, unlike passenger cars, are constructed using ECUs manufactured by different Original Equipment Manufacturers (OEMs). For reasons of interoperability, the Society of Automotive Engineers (SAE) mandates that all ECUs should communicate using the standardized SAE-J1939 protocol that gives semantics to the signals transmitted on the CAN network. Security concerns have been historically ignored in protocols and standards. Consequently, an ECU having malicious code can spoof other ECUs, e.g., a message can be injected through the OBD-II port or the telematics unit into the internal network to interfere with the behavior of the vehicle. Intrusion Detection Systems (IDS) have been proposed and utilized to detect various types of security attacks. However, such systems are only capable of detecting attacks and cannot mitigate them. A compromised ECU may generate invalid data values; even if such invalid values are detected, there is still a need to counter their effects. Almost all prior works focus on detecting attacks. We demonstrate how to make the vehicle resilient to attacks. We analyze the log files of real driving scenarios and show ECUs are significantly dependent on other ECUs to operate. We demonstrate that parameters of a compromised ECU can be reconstructed from those of other non-compromised ECUs to allow the vehicle to continue operation and make it resilient to attacks. We achieve this by modeling the behavior of an ECU using the multivariate Long Short-Term Memory (LSTM) neural network. We then reconstruct compromised ECU values using information obtained from trustworthy ECUs. Despite some levels of errors, our model can reconstruct trustworthy data values that can be substituted for values generated by compromised ECUs. The error between the reconstructed values and the correct ones is less than 6% of the operating range for the compromised ECU, which is significantly low and can be substituted. Our proposed approach makes the vehicle resilient without requiring changes to the internal architecture.

The Internet of Things (IoT) is revolutionizing society by connecting people and devices seamlessly and providing enhanced user experience and functionalities. However, the unique properties of IoT networks, such as heterogeneity and non-standardized protocol, have created critical security holes and network mismanagement. We propose a new measurement tool for IoT network data to aid in analyzing and classifying such network traffic. We use evidence from both security and machine learning research, which suggests that the complexity of a dataset can be used as a metric to determine the trustworthiness of data. We test the complexity of IoT networks using Intrinsic Dimensionality (ID), a theoretical complexity measurement based on the observation that a few variables can often describe high dimensional datasets. We use ID to evaluate four modern IoT network datasets empirically, showing that, for network and device-level data generated using IoT methodologies, the ID of the data fits into a low dimensional representation; this makes such data amenable to the use of machine learning algorithms for anomaly detection.

Abstract The dendritic cell algorithm (DCA) as one of the emerging evolutionary algorithms is based on the behavior of the specific immune agents, known as dendritic cells (DCs). DCA has several potentially beneficial features for binary classification problems. In this paper, we aim at providing a new version of this immune-inspired mechanism acts as a semi-supervised classifier, which can be a defensive shield in network intrusion detection problem. Till now, no strategy or idea has been adopted on the $Get_{Antigen()}$ function on the detection phase, but random sampling entails the DCA to provide undesirable results in several cycles at each time. This leads to uncertainty. Whereas it must be accomplished by biological behaviors of DCs in peripheral tissues, we have proposed a novel strategy that exactly acts based on its immunological functionalities of dendritic cells. The proposed mechanism focuses on two items: first, to obviate the challenge of needing to have a preordered antigen set for computing danger signal, and the second, to provide a novel immune-inspired idea for nonrandom data sampling. A variable functional migration threshold is also computed cycle by cycle that shows the necessity of the migration threshold flexibility. A significant criterion so-called capability of intrusion detection (CID) is used for tests. All the tests have been performed in a new benchmark dataset named UNSW-NB15. Experimental consequences demonstrate that the present schema as the best version among improved DC algorithms achieves 76.69% CID by 90% accuracy and outperforms its counterpart methods.

Phishing websites trick honest users into believing that they interact with a legitimate website and capture sensitive information, such as user names, passwords, credit card numbers, and other personal information. Machine learning is a promising technique to distinguish between phishing and legitimate websites. However, machine learning approaches are susceptible to adversarial learning attacks where a phishing sample can bypass classifiers. Our experiments on publicly available datasets reveal that the phishing detection mechanisms are vulnerable to adversarial learning attacks. We investigate the robustness of machine learning-based phishing detection in the face of adversarial learning attacks. We propose a practical approach to simulate such attacks by generating adversarial samples through direct feature manipulation. To enhance the sample’s success probability, we describe a clustering approach that guides an attacker to select the best possible phishing samples that can bypass the classifier by appearing as legitimate samples. We define the notion of vulnerability level for each dataset that measures the number of features that can be manipulated and the cost for such manipulation. Further, we clustered phishing samples and showed that some clusters of samples are more likely to exhibit higher vulnerability levels than others. This helps an adversary identify the best candidates of phishing samples to generate adversarial samples at a lower cost. Our finding can be used to refine the dataset and develop better learning models to compensate for the weak samples in the training dataset.

Malicious actors often use phishing attacks to compromise legitimate users' credentials. Machine learning is a promising approach for phishing detection. While the accuracy of machine learning algorithms is often dependent on the training data, very little attack data for training is available. We propose an approach for augmenting existing datasets that can be used by machine learning algorithms. We use an Adversarial Autoencoder (AAE) to generate samples that mimic the phishing websites and provide metrics to assess the quality of the generated samples. We test these samples against models trained with real-world data. Some of generated samples are able to evade existing detection model. We then use a portion of these samples in training. The new machine learning models are more robust and have higher accuracy. In other words, real-world phishing site data augmented with AAE synthesized data used for training the model is more effective for phishing detection.

Phishing websites remain a persistent security threat. Thus far, machine learning approaches appear to have the best potential as defenses. But, there are two main concerns with existing machine learning approaches for phishing detection. The first is the large number of training features used and the lack of validating arguments for these feature choices. The second concern is the type of datasets used in the literature that are inadvertently biased with respect to the features based on the website URL or content. To address these concerns, we put forward the intuition that the domain name of phishing websites is the tell-tale sign of phishing and holds the key to successful phishing detection. Accordingly, we design features that model the relationships, visual as well as statistical, of the domain name to the key elements of a phishing website, which are used to snare the end-users. The main value of our feature design is that, to bypass detection, an attacker will find it very difficult to tamper with the visual content of the phishing website without arousing the suspicion of the end user. Our feature set ensures that there is minimal or no bias with respect to a dataset. Our learning model trains with only seven features and achieves a true positive rate of 98% and a classification accuracy of 97%, on sample dataset. Compared to the state-of-the-art work, our per data instance classification is 4 times faster for legitimate websites and 10 times faster for phishing websites. Importantly, we demonstrate the shortcomings of using features based on URLs as they are likely to be biased towards specific datasets. We show the robustness of our learning algorithm by testing on unknown live phishing URLs and achieve a high detection accuracy of $99.7%$.

This paper attempts to introduce a method for developing secure software based on the vulnerabilities which are already known. In the proposed method, the most prevalent vulnerabilities are selected. For each vulnerability its location of appearance within the software development process, as well as methods of mitigation through design-level or implementation-level activities is discussed. Mapping vulnerabilities to design and implementation within software development process not only results to a better understanding of vulnerability emergence, but also allows countermeasures to be applied during initial steps of vulnerability creation, and thus better software security.This mapping shows that choosing a suitable programming language and enforcing the least privileges are the most vital design time decisions. Also, security code review and server side input validation are implementation-level activities assumed to cover most of the vulnerabilities.