Pitfalls in Machine Learning for Computer Security
Generic pitfalls related to machine learning affect all security domains and can affect the entire machine learning workflow, weakening assumptions, conclusions, and lessons learned.
Research Highlights
Advertisement
Technical Perspective: Machine Learning in Computer Security is Difficult to Fix
The study points out some common issues hindering the design of ML models for computer security and how to overcome them.
BentoMuseum: 3D and Layered Interactive Museum Map for Blind Visitors
BentoMuseum is a layered, stackable, 3D museum map that makes complex structural information accessible by allowing explorations on a floor and between floors.
Technical Perspective: An Accessible Solution to Unlock Museums
The BentoMuseum introduces a novel interactive way for visitors to understand museum layouts before their visit.
Neural Architecture Search as Program Transformation Exploration
This paper shows that program and neural architecture transformations can be interleaved delivering significant performance improvement and greater expressivity.
Technical Perspective: Optimizing Convolution Neural Nets with a Unified Transformation Approach
The key idea in "Neural Architecture Search for Program Transformation Exploration," by Jack Turner et al., is to express model architecture search as a program transformation, such that it can be naturally unified with the optimization and compilation process.
A Security Model for Web-Based Communication
We propose an algorithmic security model based on the widely deployed technologies DNSSEC and Web PKI to cover the dimensions of identification, resolution, and transaction.
Technical Perspective: Revealing the Cracks in AA Services
"A Security Model for Web-Based Communication," by Pouyan Fotouhi Tehrani et al., presents a new study of alerting authorities and their cybersafety measures.
Solving Sparse Linear Systems Faster than Matrix Multiplication
In this paper, we present an algorithm that solves linear systems with sparse coefficient matrices asymptotically faster than matrix multiplication for any ω > 2.
Technical Perspective: Solve for x
Peng and Vempala have pushed through a barrier that has stood for decades. How far will their momentum carry us?
Technical Perpective: Looking Ahead at Inclusive Technology
A generative AI technology designed to support people with dyslexia.
LaMPost: AI Writing Assistance for Adults with Dyslexia Using Large Language Models
Researchers present LaMPost, an email-writing tool with the potential to support dyslexic users.
Technical Perspective: Tracing the Network Traffic Fingerprinting Techniques of OpenVPN
"OpenVPN is Open to VPN Fingerprinting," by Diwen Xue et al., investigates network traffic fingerprinting of a very popular privacy technique—OpenVPN.
Technical Perspective: The Software-Centric Approach of SYNERGY
The advantages of a software rather than hardware approach to FPGA virtualization.
Compiler-Driven FPGA Virtualization with SYNERGY
SYNERGY virtualizes FPGAs to be used effectively in datacenters.
Technical Perspective: Creating the Internet of Biological and Bio-Inspired Things
"mSAIL: Milligram-Scale Multi-modal Sensor Platform for Monarch Butterfly Migration Tracking," by Inhee Lee et al., describes a critical milestone in the mobile system community’s efforts toward creating the Internet of biological things.
mSAIL: Milligram-Scale Multi-Modal Sensor Platform for Monarch Butterfly Migration Tracking
We propose a wireless sensing platform, mSAIL, specifically designed for a monarch butterfly migration study based on previously developed custom-designed ICs.
Dynamic Placement in Refugee Resettlement
In this paper, we design an online algorithm for refugee allocation.
Technical Perspective: Improving Refugees’ Integration with Online Resource Allocation
"Dynamic Placement in Refugee Resettlement," by Narges Ahani et al., applies online resource allocation concepts to a new and important area: refugee resettlement.
Combining Machine Learning and Lifetime-Based Resource Management for Memory Allocation and Beyond
We introduce a two-step approach to attain high memory utilization in huge pages, which gives rise to a new methodology for applying Machine Learning in computer systems.
Technical Perspective: Learning-Based Memory Allocation for C++ Server Workloads
"Learning and Lifetime-Based Resource Management for Memory Allocation and Beyond," by Martin Maas et al., explores the potential of using imperfect information in the design of memory managers.
Indistinguishability Obfuscation from Well-Founded Assumptions
We examine a formalization of the “one-way compiler" concept with the notion of indistinguishability obfuscation.
Technical Perspective: Hiding Secrets in Programs
"Indistinguishability Obfuscation from Well-Founded Assumptions," by Aayush Jain et al., gives a new construction of indistinguishability obfuscation that is provably secure.
Taming Algorithmic Priority Inversion in Mission-Critical Perception Pipelines
This paper discusses algorithmic priority inversion in mission-critical machine inference pipelines used in modern neural-network-based perception subsystems and describes a solution to mitigate its effect.
Shape the Future of Computing
ACM encourages its members to take a direct hand in shaping the future of the association. There are more ways than ever to get involved.
Get InvolvedCommunications of the ACM (CACM) is now a fully Open Access publication.
By opening CACM to the world, we hope to increase engagement among the broader computer science community and encourage non-members to discover the rich resources ACM has to offer.
Learn More