Fine-Tuning Llama 2 Large Language Models for Detecting Online Sexual Predatory Chats and Abusive Texts

Detecting online sexual predatory behaviours and abusive language on social media platforms has become a critical area of research due to the growing concerns about online safety, especially for vulnerable populations such as children and adolescents. Researchers have been exploring various techniques and approaches to develop effective detection systems that can identify and mitigate these risks. Recent development of large language models (LLMs) has opened a new opportunity to address this problem more effectively. This paper proposes an approach to detection of online sexual predatory chats and abusive language using the open-source pretrained Llama 2 7B-parameter model, recently released by Meta GenAI. We fine-tune the LLM using datasets with different sizes, imbalance degrees, and languages (i.e., English, Roman Urdu and Urdu). Based on the power of LLMs, our approach is generic and automated without a manual search for a synergy between feature extraction and classifier design steps like conventional methods in this domain. Experimental results show a strong performance of the proposed approach, which performs proficiently and consistently across three distinct datasets with five sets of experiments. This study's outcomes indicate that the proposed method can be implemented in real-world applications (even with non-English languages) for flagging sexual predators, offensive or toxic content, hate speech, and discriminatory language in online discussions and comments to maintain respectful internet or digital communities. Furthermore, it can be employed for solving text classification problems with other potential applications such as sentiment analysis, spam and phishing detection, sorting legal documents, fake news detection, language identification, user intent recognition, text-based product categorization, medical record analysis, and resume screening.

The Data Airlock: infrastructure for restricted data informatics

Data science collaboration is problematic when access to operational data or models from outside the data-holding organisation is prohibited, for a variety of legal, security, ethical, or practical reasons. There are significant data privacy challenges when performing collaborative data science work against such restricted data. In this paper we describe a range of causes and risks associated with restricted data along with the social, environmental, data, and cryptographic measures that may be used to mitigate such issues. We then show how these are generally inadequate for restricted data contexts and introduce the 'Data Airlock' - secure infrastructure that facilitates 'eyes-off' data science workloads. After describing our use-case we detail the architecture and implementation of a first, single-organisation version of the Data Airlock infrastructure. We conclude with outcomes and learning from this implementation, and outline requirements for a second, federated version.

PDQ & TMK + PDQF -- A Test Drive of Facebook's Perceptual Hashing Algorithms

Efficient and reliable automated detection of modified image and multimedia files has long been a challenge for law enforcement, compounded by the harm caused by repeated exposure to psychologically harmful materials. In August 2019 Facebook open-sourced their PDQ and TMK + PDQF algorithms for image and video similarity measurement, respectively. In this report, we review the algorithms' performance on detecting commonly encountered transformations on real-world case data, sourced from contemporary investigations. We also provide a reference implementation to demonstrate the potential application and integration of such algorithms within existing law enforcement systems.