The usage of deep neural networks in safety-critical systems is limited by our ability to guarantee their correct behavior. Runtime monitors are components aiming to identify unsafe predictions and discard them before they can lead to catastrophic consequences. Several recent works on runtime monitoring have focused on out-of-distribution (OOD) detection, i.e., identifying inputs that are different from the training data. In this work, we argue that OOD detection is not a well-suited framework to design efficient runtime monitors and that it is more relevant to evaluate monitors based on their ability to discard incorrect predictions. We call this setting out-ofmodel-scope detection and discuss the conceptual differences with OOD. We also conduct extensive experiments on popular datasets from the literature to show that studying monitors in the OOD setting can be misleading: 1. very good OOD results can give a false impression of safety, 2. comparison under the OOD setting does not allow identifying the best monitor to detect errors. Finally, we also show that removing erroneous training data samples helps to train better monitors.

Machine learning applications, especially in the fields of me\-di\-cine and social sciences, are slowly being subjected to increasing scrutiny. Similarly to sample size planning performed in clinical and social studies, lawmakers and funding agencies may expect statistical uncertainty estimations in machine learning applications that impact society. In this paper, we present an easy-to-use python package and web application for estimating prediction confidence intervals. The package offers eight different procedures to determine and justify the sample size and confidence of predictions from holdout, bootstrap, cross-validation, and progressive validation experiments. Since the package builds directly on established data analysis libraries, it seamlessly integrates into preprocessing and exploratory data analysis steps. Code related to this paper is available at: https://github.com/dabrze/confidence-planner.

An important aspect of many botnets is their capability to generate pseudorandom domain names using Domain Generation Algorithms (DGAs). A cyber criminal can register such domains to establish periodically changing rendezvous points with the bots. DGAs make use of seeds to generate sets of domains. Seeds can easily be changed in order to generate entirely new groups of domains while using the same underlying algorithm. While this requires very little manual effort for an adversary, security specialists typically have to manually reverse engineer new malware strains to reconstruct the seeds. Only when the seed and DGA are known, past and future domains can be generated, efficiently attributed, blocked, sinkholed or used for a take-down. Common counters in the literature consist of databases or Machine Learning (ML) based detectors to keep track of past and future domains of known DGAs and to identify DGA-generated domain names, respectively. However, database based approaches can not detect domains generated by new DGAs, and ML approaches can not generate future domain names. In this paper, we introduce SESAME, a system that combines the two above-mentioned approaches and contains a module for automatic Seed Reconstruction, which is, to our knowledge, the first of its kind. It is used to automatically classify domain names, rate their novelty, and determine the seeds of the underlying DGAs. SESAME consists of multiple DGA-specific Seed Reconstructors and is designed to work purely based on domain names, as they are easily obtainable from observing the network traffic. We evaluated our approach on 20.8 gigabytes of DNS-lookups. Thereby, we identified 17 DGAs, of which 4 were entirely new to us.

Blockchain technology has been successfully exploited for deploying new economic applications. However, it has started arousing the interest of malicious users who deliver scams to deceive honest users and to gain economic advantages. Among the various scams, Ponzi schemes are one of the most common. Here, we present an automatic technique for detecting smart Ponzi contracts on Ethereum. We release a reusable data set with 4422 unique real-world smart contracts. Then, we introduce a new set of features that allow us to improve the classification. Finally, we identify a small and effective set of features that ensures a good classification quality.

We show, to our knowledge, the first theoretical treatments of two common questions in cross-validation based hyperparameter selection: (1) After selecting the best hyperparameter using a held-out set, we train the final model using {\em all} of the training data -- since this may or may not improve future generalization error, should one do this? (2) During optimization such as via SGD (stochastic gradient descent), we must set the optimization tolerance $\rho$ -- since it trades off predictive accuracy with computation cost, how should one set it? Toward these problems, we introduce the {\em hold-in risk} (the error due to not using the whole training data), and the {\em model class mis-specification risk} (the error due to having chosen the wrong model class) in a theoretical view which is simple, general, and suggests heuristics that can be used when faced with a dataset instance. In proof-of-concept studies in synthetic data where theoretical quantities can be controlled, we show that these heuristics can, respectively, (1) always perform at least as well as always performing retraining or never performing retraining, (2) either improve performance or reduce computational overhead by $2\times$ with no loss in predictive performance.

Predicting the novel effects of drugs based on information about approved drugs can be regarded as a recommendation system. Matrix factorization is one of the most used recommendation systems and various algorithms have been devised for it. A literature survey and summary of existing algorithms for predicting drug effects demonstrated that most such methods, including neighborhood regularized logistic matrix factorization, which was the best performer in benchmark tests, used a binary matrix that considers only the presence or absence of interactions. However, drug effects are known to have two opposite aspects, such as side effects and therapeutic effects. In the present study, we proposed using neighborhood regularized bidirectional matrix factorization (NRBdMF) to predict drug effects by incorporating bidirectionality, which is a characteristic property of drug effects. We used this proposed method for predicting side effects using a matrix that considered the bidirectionality of drug effects, in which known side effects were assigned a positive label (plus 1) and known treatment effects were assigned a negative (minus 1) label. The NRBdMF model, which utilizes drug bidirectional information, achieved enrichment of side effects at the top and indications at the bottom of the prediction list. This first attempt to consider the bidirectional nature of drug effects using NRBdMF showed that it reduced false positives and produced a highly interpretable output.

Out-of-Distribution (OOD) detection, i.e., identifying whether an input is sampled from a novel distribution other than the training distribution, is a critical task for safely deploying machine learning systems in the open world. Recently, post hoc detection utilizing pre-trained models has shown promising performance and can be scaled to large-scale problems. This advance raises a natural question: Can we leverage the diversity of multiple pre-trained models to improve the performance of post hoc detection methods? In this work, we propose a detection enhancement method by ensembling multiple detection decisions derived from a zoo of pre-trained models. Our approach uses the p-value instead of the commonly used hard threshold and leverages a fundamental framework of multiple hypothesis testing to control the true positive rate of In-Distribution (ID) data. We focus on the usage of model zoos and provide systematic empirical comparisons with current state-of-the-art methods on various OOD detection benchmarks. The proposed ensemble scheme shows consistent improvement compared to single-model detectors and significantly outperforms the current competitive methods. Our method substantially improves the relative performance by 65.40% and 26.96% on the CIFAR10 and ImageNet benchmarks.

In the research area of anomaly detection, novel and promising methods are frequently developed. However, most existing studies exclusively focus on the detection task only and ignore the interpretability of the underlying models as well as their detection results. Nevertheless, anomaly interpretation, which aims to provide explanation of why specific data instances are identified as anomalies, is an equally important task in many real-world applications. In this work, we propose a novel framework which synergizes several machine learning and data mining techniques to automatically learn invariant rules that are consistently satisfied in a given dataset. The learned invariant rules can provide explicit explanation of anomaly detection results in the inference phase and thus are extremely useful for subsequent decision-making regarding reported anomalies. Furthermore, our empirical evaluation shows that the proposed method can also achieve comparable or even better performance in terms of AUC and partial AUC on public benchmark datasets across various application domains compared with start-of-the-art anomaly detection models.

Originally published on Towards AI. I used an open-sourced library, cleanlab, to remove low-quality labels on an image dataset. The model trained on the dataset without low-quality data gained 4 percentage points of accuracy compared to the baseline model (trained on all data). Improving data quality sounds easy enough. But the workload of manually checking data quality can quickly become insurmountable as the dataset scales.

L2 regularization is a method used to prevent overfitting in machine learning models. It adds a penalty term to the loss function that is proportional to the sum of the squares of the weights. This penalizes large weights and encourages the model to use only the most relevant features. L2 regularization is also known as weight decay because it causes the model's weight to decay toward zero. The penalty term is usually added to the loss function during training and is typically set by cross-validation.