Large language models (LLMs) often exhibit undesirable behaviors, such as safety violations and hallucinations.Although inference-time steering offers a cost-effective way to adjust model behavior without updating its parameters, existing methods often fail to be simultaneously effective, utilitypreserving, and training-efficient due to their rigid, one-size-fits-all designs and limited adaptability.In this work, we present FineSteer, a novel steering framework that decomposes inference-time steering into two complementary stages-conditional steering and fine-grained vector synthesis-allowing finegrained control over when and how to steer internal representations.In the first stage, we introduce a Subspace-guided Conditional Steering (SCS) mechanism that preserves model utility by avoiding unnecessary steering.In the second stage, we propose a Mixture-of-Steering-Experts (MoSE) mechanism that captures the multimodal nature of desired steering behaviors and generates query-specific steering vectors for improved effectiveness.Through tailored designs in both SCS and MoSE, FineSteer maintains robust performance on general queries while adaptively optimizing steering vectors for targeted inputs in a training-efficient manner.Extensive experiments on safety and truthfulness benchmarks show that FineSteer outperforms the state-of-the-art methods in overall performance (e.g., A 7.6% improvement on TruthfulQA over Llama-3), achieving stronger steering performance with minimal utility loss.The code is available at https://github.com/YukinoAsuna/FineSteer.

Smart-home automations enable users to customize smart devices to react automatically to people, the environment, and more. For example, an automation might adjust the lights when people are at home or enable a garage door to open by voice command. While automations offer convenience and accessibility, they can also inadvertently expose users to security and privacy risks, such as leaking sensitive data or allowing untrusted parties to control users' devices. Prior work has shown that information flow analysis is a promising technique for identifying these kinds of risks, hypothesizing that the analysis would be yet more effective if it could differentiate between devices located in different places in the home. We tested this hypothesis by developing a tool that extends prior information flow analysis approaches to account for device location. We conducted an interview study with 22 participants to build a dataset of home automations to establish a ground truth to evaluate the tool. We found that incorporating device location leads to an improved analysis that identifies more of the vulnerabilities users care about (F1 score 0.74) compared to prior work (F1 score 0.29). Our results demonstrate the feasibility of incorporating device location into an information flow analysis and, perhaps more importantly, suggest additional ways to prevent security and privacy risks beyond controlling potentially unsafe information flows.

In this study, we reveal, for the first time, popular online social networks (especially Twitter) are being extensively abused by miscreants to promote illicit goods and services of diverse categories. This study is made possible by multiple machine learning tools that are designed to detect and analyze Posts of Illicit Promotion (PIPs) as well as revealing their underlying promotion campaigns. Particularly, we observe that PIPs are prevalent on Twitter, along with extensive visibility on other three popular OSNs including YouTube, Facebook, and TikTok. For instance, applying our PIP hunter to the Twitter platform for 6 months has led to the discovery of 12 million distinct PIPs which are widely distributed in 5 major natural languages and 10 illicit categories, e.g., drugs, data leakage, gambling, and weapon sales. Along the discovery of PIPs are 580K Twitter accounts publishing PIPs as well as 37K distinct instant messaging accounts that are embedded in PIPs and serve as next hops of communication with prospective customers. Also, an arms race between Twitter and illicit promotion operators is also observed. Especially, 90% PIPs can survice the first two months since getting published on Twitter, which is likely due to the diverse evasion tactics adopted by miscreants to masquerade PIPs.