Ping Storms at GreyNoise
Comments
Month: September 2024
Fidelity has cut X’s value by 79% since Musk purchase
Elon Musk’s X is now valued at less than a quarter of its $44 billion purchase price, according to a new estimate from investor Fidelity. The asset manager, which helped Musk acquire the social network formerly known as Twitter, has further reduced the value of its holding in X to a total markdown of 78.7% … Read more
Bitcoin bull run unlikely as long as social sentiment runs too hot — Santiment
Santiment says there’s nearly double the amount of bullish posts to bearish ones on social media.
Ethereum Staking Yields Could Outpace US Rates, Boosting Prices Experts Say
A “double-whammy effect” is expected to narrow the gap between ETH staking returns and traditional risk-free rates in the coming quarters.
How Blockchain APIs and RPC Nodes Work
Ironically, a lot of developers and projects still significantly depend on centralized services in the area of blockchain, where decentralization rules. The foundation of blockchain-based apps (or dApps) and services is made up of blockchain APIs and RPC nodes in this situation. With the help of these tools, developers may communicate with blockchain networks without … Read more
California governor vetoes hotly-contested AI safety bill
Gavin Newsom has vetoed SB 1048, saying that “while well-intentioned,” it could place unnecessary restrictions on emerging AI companies in California.
The Ultimate Oldschool PC Font Pack
Comments
Vitalik Buterin proposes metrics to measure ‘Ethereum alignment’
Ethereum-based projects should have metrics to strive toward to ensure they are collectively “building something that feels like one Ethereum ecosystem.”
Meta’s Software Engineer Levels Explained
Meta Levels for Software Engineers Meta (formerly Facebook) has a well-defined structure and levels for software engineers, starting from E2 (IC2) for interns/students. This article will focus on levels E3 (IC3) to E10 (IC10), covering fully qualified engineers and individual contributors. These levels define career progression, responsibilities, and compensation, ensuring clarity and growth opportunities for … Read more
Enhancing Kubernetes Resource Management with SharedInformers
One of the roles of the Kubernetes controller is to monitor objects for the desired vs actual states and then send requests to change the existing state to the desired state at a controlled rate. The controller requests the Kubernetes API server to retrieve an object’s information. However, frequently retrieving data from the API server … Read more
Crypto wallet drainer was on Google Play for months, stole $70K: Report
The malicious wallet-draining app marked “the first time drainers exclusively targeted mobile users,” says Check Point Research.
Comparison with SKD and ARD and Implementations of Stronger Attacker Algorithms
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Evaluating NEO-KD Against Single-Exit Defense Methods in Multi-Exit Networks
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Examining the Adversarial Test Accuracy of Later Exits in NEO-KD Networks
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
The Impact of Hyperparameters on Adversarial Training Performance
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Clean Test Accuracy and Adversarial Training via Average Attack
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Fine-Tuning NEO-KD for Robust Multi-Exit Networks
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
How NEO-KD Reduces Adversarial Transferability and Improves Accuracy
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
How Ensemble Strategies Impact Adversarial Robustness in Multi-Exit Networks
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
How NEO-KD Saves Up to 81% of Computing Power While Maximizing Adversarial Accuracy
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Benchmarking NEO-KD on Adversarial Robustness
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
A Robust Self-Distillation Strategy for Multi-Exit Networks
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Adversarial Training in Multi-Exit Networks: Proposed NEO-KD Algorithm and Problem Setup
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more
Advancing Robustness in Multi-Exit Networks Through Exit-Wise Knowledge Distillation
:::info Authors: (1) Seokil Ham, KAIST; (2) Jungwuk Park, KAIST; (3) Dong-Jun Han, Purdue University; (4) Jaekyun Moon, KAIST. ::: Table of Links Abstract and 1. Introduction 2. Related Works 3. Proposed NEO-KD Algorithm and 3.1 Problem Setup: Adversarial Training in Multi-Exit Networks 3.2 Algorithm Description 4. Experiments and 4.1 Experimental Setup 4.2. Main Experimental … Read more