Feature: Software and tools
information is used as part of the prompt to give the LLM better context to craſt a complete and accurate solution.
Security and privacy Off-premises LLMs pose significant security and privacy challenges, particularly when organisations are required to share sensitive or proprietary information. When data is transmitted to and processed by external cloud-based LLM services, there is a risk that confidential information, such as trade secrets, intellectual property or customer data, may be stored or logged by the service provider. This can lead to unintended exposure if the provider’s security measures are inadequate or if there is a breach in the cloud infrastructure. Consequently, organisations must ensure that data encryption is used both in transit and at rest, and that the third-party vendor adheres to strict data protection regulations and industry best practices. In addition to privacy concerns, there are considerable security
risks associated with off-premises LLMs. Te possibility of adversarial attacks, where malicious actors manipulate prompts or exploit vulnerabilities in the model’s architecture, can lead to unauthorised data access or even compromise the integrity of the model’s outputs. Furthermore, the opaque nature of some cloud-based LLM services may make it difficult to ascertain how
data is handled, stored or, potentially, repurposed. To mitigate these risks, organisations should conduct thorough security assessments of potential LLM providers, establish robust access controls and consider hybrid or local on-premises deployment of LLMs for processing highly sensitive information. Tese measures help safeguard confidential data and ensure compliance with relevant security and privacy standards.
Producing better fixes Manual inspection of the sample data revealed that fixes generated with the AI-augmented commercial engine’s prompts are oſten more complete (e.g., fixing all instances of an issue in adjacent lines), robust (better error handling) and conform to standard practices. We believe that including rule documentation and enforcing chain-of-thought reasoning in the prompts – either through reasoning questions or the structure of bare prompts – helps the model produce better fixes. We see that using an LLM saves time in understanding a
coding violation as well as time in craſting a fix. Additionally, it’s clear that extra effort and information produce a better prompt, which will end up saving time and producing better code. Proper adoption of LLMs as a part of the soſtware compliance strategy can help handle the increasing volume of code in today’s modern vehicles.
Magazine app
www.electronicsworld.co.uk
February 2025 Volume 130 Issue 2043 £5.90
tap
IN THIS ISSUE: • Sensor Technology • Power • Display • Memory • RF Design
the
Search for Electronics World magazine on the App Store and Google Play
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48
