Konstantin Berlin

Konstantin Berlin, Adam Swanda

Many automated labeling pipelines classify inputs into categories defined by a written specification, content moderation being a prominent use case. Simple category definitions are not detailed enough for labelers to produce the accurate, consistent golden labels these pipelines require. One solution is to write a prescriptive definition that settles enough real boundary cases that labelers cannot disagree with the written interpretation. In practice, definitions at that level of detail exceed what a human annotator can hold in working memory, so annotators fall back on intuition and the labels drift from the written rules, regressing on accuracy and consistency. We propose and demonstrate the efficacy of an AI-driven workflow in which AI helps write a per-category constitution that defines the label in enough detail to cover edge cases, and a frontier LLM interprets it on each input to produce the golden label more consistently and accurately than humans reading the same document. We evaluate on three content moderation categories (harassment, hate speech, non-violent crime) and show that the approach reduces cross-model inconsistency by up to 57x compared to paragraph definitions, with cross-model disagreement diagnosing specification gaps and the human responsible for high-level decisions about what each category should mean rather than individual labeling calls. For the safety evaluation, we introduce a dual-axis formulation scoring intent and content independently over the full conversation, so downstream consumers can act on either axis or both.

Ben Gelman, Salma Taoufiq, Tamás Vörös et al.

In place of in-house solutions, organizations are increasingly moving towards managed services for cyber defense. Security Operations Centers are specialized cybersecurity units responsible for the defense of an organization, but the large-scale centralization of threat detection is causing SOCs to endure an overwhelming amount of false positive alerts -- a phenomenon known as alert fatigue. Large collections of imprecise sensors, an inability to adapt to known false positives, evolution of the threat landscape, and inefficient use of analyst time all contribute to the alert fatigue problem. To combat these issues, we present That Escalated Quickly (TEQ), a machine learning framework that reduces alert fatigue with minimal changes to SOC workflows by predicting alert-level and incident-level actionability. On real-world data, the system is able to reduce the time it takes to respond to actionable incidents by $22.9\%$, suppress $54\%$ of false positives with a $95.1\%$ detection rate, and reduce the number of alerts an analyst needs to investigate within singular incidents by $14\%$.

Adam Swanda, Amy Chang, Alexander Chen et al.

The widespread adoption of Large Language Models (LLMs) has revolutionized AI deployment, enabling autonomous and semi-autonomous applications across industries through intuitive language interfaces and continuous improvements in model development. However, the attendant increase in autonomy and expansion of access permissions among AI applications also make these systems compelling targets for malicious attacks. Their inherent susceptibility to security flaws necessitates robust defenses, yet no known approaches can prevent zero-day or novel attacks against LLMs. This places AI protection systems in a category similar to established malware protection systems: rather than providing guaranteed immunity, they minimize risk through enhanced observability, multi-layered defense, and rapid threat response, supported by a threat intelligence function designed specifically for AI-related threats. Prior work on LLM protection has largely evaluated individual detection models rather than end-to-end systems designed for continuous, rapid adaptation to a changing threat landscape. We present a production-grade defense system rooted in established malware detection and threat intelligence practices. Our platform integrates three components: a threat intelligence system that turns emerging threats into protections; a data platform that aggregates and enriches information while providing observability, monitoring, and ML operations; and a release platform enabling safe, rapid detection updates without disrupting customer workflows. Together, these components deliver layered protection against evolving LLM threats while generating training data for continuous model improvement and deploying updates without interrupting production.

Tamás Vörös, Sean Paul Bergeron, Konstantin Berlin

We introduce a state-of-the-art approach for URL categorization that leverages the power of Large Language Models (LLMs) to address the primary objectives of web content filtering: safeguarding organizations from legal and ethical risks, limiting access to high-risk or suspicious websites, and fostering a secure and professional work environment. Our method utilizes LLMs to generate accurate classifications and then employs established knowledge distillation techniques to create smaller, more specialized student models tailored for web content filtering. Distillation results in a student model with a 9% accuracy rate improvement in classifying websites, sourced from customer telemetry data collected by a large security vendor, into 30 distinct content categories based on their URLs, surpassing the current state-of-the-art approach. Our student model matches the performance of the teacher LLM with 175 times less parameters, allowing the model to be used for in-line scanning of large volumes of URLs, and requires 3 orders of magnitude less manually labeled training data than the current state-of-the-art approach. Depending on the specific use case, the output generated by our approach can either be directly returned or employed as a pre-filter for more resource-intensive operations involving website images or HTML.

Awalin Sopan, Konstantin Berlin

Development of new machine learning models is typically done on manually curated data sets, making them unsuitable for evaluating the models' performance during operations, where the evaluation needs to be performed automatically on incoming streams of new data. Unfortunately, pure reliance on a fully automatic pipeline for monitoring model performance makes it difficult to understand if any observed performance issues are due to model performance, pipeline issues, emerging data distribution biases, or some combination of the above. With this in mind, we developed a web-based visualization system that allows the users to quickly gather headline performance numbers while maintaining confidence that the underlying data pipeline is functioning properly. It also enables the users to immediately observe the root cause of an issue when something goes wrong. We introduce a novel way to analyze performance under data issues using a data coverage equalizer. We describe the various modifications and additional plots, filters, and drill-downs that we added on top of the standard evaluation metrics typically tracked in machine learning (ML) applications, and walk through some real world examples that proved valuable for introspecting our models.

Konstantin Berlin, Ajay Lakshminarayanarao

Generating up to date, well labeled datasets for machine learning (ML) security models is a unique engineering challenge, as large data volumes, complexity of labeling, and constant concept drift makes it difficult to generate effective training datasets. Here we describe a simple, resilient cloud infrastructure for generating ML training and testing datasets, that has enhanced the speed at which our team is able to research and keep in production a multitude of security ML models.

Adarsh Kyadige, Ethan M. Rudd, Konstantin Berlin

Machine learning (ML) used for static portable executable (PE) malware detection typically employs per-file numerical feature vector representations as input with one or more target labels during training. However, there is much orthogonal information that can be gleaned from the \textit{context} in which the file was seen. In this paper, we propose utilizing a static source of contextual information -- the path of the PE file -- as an auxiliary input to the classifier. While file paths are not malicious or benign in and of themselves, they do provide valuable context for a malicious/benign determination. Unlike dynamic contextual information, file paths are available with little overhead and can seamlessly be integrated into a multi-view static ML detector, yielding higher detection rates at very high throughput with minimal infrastructural changes. Here we propose a multi-view neural network, which takes feature vectors from PE file content as well as corresponding file paths as inputs and outputs a detection score. To ensure realistic evaluation, we use a dataset of approximately 10 million samples -- files and file paths from user endpoints of an actual security vendor network. We then conduct an interpretability analysis via LIME modeling to ensure that our classifier has learned a sensible representation and see which parts of the file path most contributed to change in the classifier's score. We find that our model learns useful aspects of the file path for classification, while also learning artifacts from customers testing the vendor's product, e.g., by downloading a directory of malware samples each named as their hash. We prune these artifacts from our test dataset and demonstrate reductions in false negative rate of 32.3% at a $10^{-3}$ false positive rate (FPR) and 33.1% at $10^{-4}$ FPR, over a similar topology single input PE file content only model.

Felipe N. Ducau, Ethan M. Rudd, Tad M. Heppner et al.

With the rapid proliferation and increased sophistication of malicious software (malware), detection methods no longer rely only on manually generated signatures but have also incorporated more general approaches like machine learning detection. Although powerful for conviction of malicious artifacts, these methods do not produce any further information about the type of threat that has been detected neither allows for identifying relationships between malware samples. In this work, we address the information gap between machine learning and signature-based detection methods by learning a representation space for malware samples in which files with similar malicious behaviors appear close to each other. We do so by introducing a deep learning based tagging model trained to generate human-interpretable semantic descriptions of malicious software, which, at the same time provides potentially more useful and flexible information than malware family names. We show that the malware descriptions generated with the proposed approach correctly identify more than 95% of eleven possible tag descriptions for a given sample, at a deployable false positive rate of 1% per tag. Furthermore, we use the learned representation space to introduce a similarity index between malware files, and empirically demonstrate using dynamic traces from files' execution, that is not only more effective at identifying samples from the same families, but also 32 times smaller than those based on raw feature vectors.

Ethan M. Rudd, Felipe N. Ducau, Cody Wild et al.

Malware detection is a popular application of Machine Learning for Information Security (ML-Sec), in which an ML classifier is trained to predict whether a given file is malware or benignware. Parameters of this classifier are typically optimized such that outputs from the model over a set of input samples most closely match the samples' true malicious/benign (1/0) target labels. However, there are often a number of other sources of contextual metadata for each malware sample, beyond an aggregate malicious/benign label, including multiple labeling sources and malware type information (e.g., ransomware, trojan, etc.), which we can feed to the classifier as auxiliary prediction targets. In this work, we fit deep neural networks to multiple additional targets derived from metadata in a threat intelligence feed for Portable Executable (PE) malware and benignware, including a multi-source malicious/benign loss, a count loss on multi-source detections, and a semantic malware attribute tag loss. We find that incorporating multiple auxiliary loss terms yields a marked improvement in performance on the main detection task. We also demonstrate that these gains likely stem from a more informed neural network representation and are not due to a regularization artifact of multi-target learning. Our auxiliary loss architecture yields a significant reduction in detection error rate (false negatives) of 42.6% at a false positive rate (FPR) of $10^{-3}$ when compared to a similar model with only one target, and a decrease of 53.8% at $10^{-5}$ FPR.

Joshua Saxe, Konstantin Berlin

For years security machine learning research has promised to obviate the need for signature based detection by automatically learning to detect indicators of attack. Unfortunately, this vision hasn't come to fruition: in fact, developing and maintaining today's security machine learning systems can require engineering resources that are comparable to that of signature-based detection systems, due in part to the need to develop and continuously tune the "features" these machine learning systems look at as attacks evolve. Deep learning, a subfield of machine learning, promises to change this by operating on raw input signals and automating the process of feature design and extraction. In this paper we propose the eXpose neural network, which uses a deep learning approach we have developed to take generic, raw short character strings as input (a common case for security inputs, which include artifacts like potentially malicious URLs, file paths, named pipes, named mutexes, and registry keys), and learns to simultaneously extract features and classify using character-level embeddings and convolutional neural network. In addition to completely automating the feature design and extraction process, eXpose outperforms manual feature extraction based baselines on all of the intrusion detection problems we tested it on, yielding a 5%-10% detection rate gain at 0.1% false positive rate compared to these baselines.

Konstantin Berlin, Joshua Saxe

Enterprise networks are in constant danger of being breached by cyber-attackers, but making the decision about what security tools to deploy to mitigate this risk requires carefully designed evaluation of security products. One of the most important metrics for a protection product is how well it is able to stop malware, specifically on "zero"-day malware that has not been seen by the security community before. However, evaluating zero-day performance is difficult, because of larger number of previously unseen samples that are needed to properly measure the true and false positive rate, and the challenges involved in accurately labeling these samples. This paper addresses these issues from a statistical and practical perspective. Our contributions include first showing that the number of benign files needed for proper evaluation is on the order of a millions, and the number of malware samples needed is on the order of tens of thousands. We then propose and justify a time-delay method for easily collecting large number of previously unseen, but labeled, samples. This enables cheap and accurate evaluation of zero-day true and false positive rates. Finally, we propose a more fine-grain labeling of the malware/benignware in order to better model the heterogeneous distribution of files on various networks.

Joshua Saxe, Konstantin Berlin

Malware remains a serious problem for corporations, government agencies, and individuals, as attackers continue to use it as a tool to effect frequent and costly network intrusions. Machine learning holds the promise of automating the work required to detect newly discovered malware families, and could potentially learn generalizations about malware and benign software that support the detection of entirely new, unknown malware families. Unfortunately, few proposed machine learning based malware detection methods have achieved the low false positive rates required to deliver deployable detectors. In this paper we a deep neural network malware classifier that achieves a usable detection rate at an extremely low false positive rate and scales to real world training example volumes on commodity hardware. Specifically, we show that our system achieves a 95% detection rate at 0.1% false positive rate (FPR), based on more than 400,000 software binaries sourced directly from our customers and internal malware databases. We achieve these results by directly learning on all binaries, without any filtering, unpacking, or manually separating binary files into categories. Further, we confirm our false positive rates directly on a live stream of files coming in from Invincea's deployed endpoint solution, provide an estimate of how many new binary files we expected to see a day on an enterprise network, and describe how that relates to the false positive rate and translates into an intuitive threat score. Our results demonstrate that it is now feasible to quickly train and deploy a low resource, highly accurate machine learning classification model, with false positive rates that approach traditional labor intensive signature based methods, while also detecting previously unseen malware.

Konstantin Berlin, David Slater, Joshua Saxe

As antivirus and network intrusion detection systems have increasingly proven insufficient to detect advanced threats, large security operations centers have moved to deploy endpoint-based sensors that provide deeper visibility into low-level events across their enterprises. Unfortunately, for many organizations in government and industry, the installation, maintenance, and resource requirements of these newer solutions pose barriers to adoption and are perceived as risks to organizations' missions. To mitigate this problem we investigated the utility of agentless detection of malicious endpoint behavior, using only the standard build-in Windows audit logging facility as our signal. We found that Windows audit logs, while emitting manageable sized data streams on the endpoints, provide enough information to allow robust detection of malicious behavior. Audit logs provide an effective, low-cost alternative to deploying additional expensive agent-based breach detection systems in many government and industrial settings, and can be used to detect, in our tests, 83% percent of malware samples with a 0.1% false positive rate. They can also supplement already existing host signature-based antivirus solutions, like Kaspersky, Symantec, and McAfee, detecting, in our testing environment, 78% of malware missed by those antivirus systems.