Awarded to Qumodo Ltd

Start date: Sunday 31 March 2019
Value: £1,300,000
Company size: SME
Home Office and Policing

CAID Digital Media Examiner (DME)

3 Incomplete applications

1 SME, 2 large

9 Completed applications

6 SME, 3 large

Important dates

Published
Monday 21 January 2019
Deadline for asking questions
Monday 28 January 2019 at 11:59pm GMT
Closing date for applications
Monday 4 February 2019 at 11:59pm GMT

Overview

Summary of the work
The DME will compare the images against current known images.
Unknown images will be run through a UKLE auto-categoriser algorithm.
All legal images will then be run through image-similarity software. These will need to be integrated into the DME. Images will be exported into the Child Abuse Image Database (CAID).
Latest start date
Monday 1 April 2019
Expected contract length
24 months
Location
London
Organisation the work is for
Home Office and Policing
Budget range
Up to the maximum value of £1.3M over the term.

About the work

Why the work is being done
The Home Office is looking to improve UKLE (UK Law Enforcement) officer welfare when reviewing distressing child abuse imagery, whilst improving time in categorising illegal imagery using the latest technology in assisted categorisation.
Problem to be solved
Online Child Sexual Abuse/Exploitation (OCSAE) Officers must look through vast quantities of media (images, video, text, etc.) on multiple devices, with the intent of categorisation to deliver operational value and context. This is a very resource heavy activity (regarding time and effort) which could include facial, object & relationship matching. These images are distressing to review, and officers need the correct tools to support them in this activity.
Who the users are and what they need to do
• UKLE officers working within digital forensics and on Online Child Sexual Abuse and Exploitation cases
• Up to 50 force organisations across the UK (UKLE and other organisations within the OSCAE space).
Early market engagement
none
Any work that’s already been done
A Home Office IIoC auto-categoriser has been developed and is available for plug-in via API. Any other categoriser will not be considered.
Output definitions have been standardised and are available as Project Vic ODATA CAID compliant json files.
Existing team
The supplier will be working with the CAID PPPT project team.
Current phase
Not applicable

Work setup

Address where the work will take place
The Home Office, Marsham Street, London for project team.
Met Police site(s) available for acceptance testing and implementation.
The DME tool will be used by up to 50 UKLE Forces nationally (full list to be provided as part of the contract).
Working arrangements
Statements of Work will have specific working arrangements and base locations included.
Given sensitive nature of imagery no direct access will be given to the images. Testing will be via specific Home Office users and at a designated site.
Security clearance
All staff must hold or be prepared to undergo SC Security Clearance to operate on this project.

Additional information

Additional terms and conditions
Standard DOS3 terms and conditions however any IP generated as a result of projects commissioned by the authority, shall vest with the authority.

Skills and experience

Buyers will use the essential and nice-to-have skills and experience to help them evaluate suppliers’ technical competence.

Essential skills and experience
  • Knowledge of Image EXIF data
  • Knowledge of image types
  • Knowledge of Image Hashing
  • Knowledge of Image Matching
  • Knowledge of the use of Imagery within Police Investigations with relevance given to online CSE cases
  • UI development skills
  • Software development skills
  • Product ready for Implementation Q2 2019 (Calendar Year)
  • Must have a product that is near operational use against the requirements specified in the problem with minimal development and configuration needed
  • Development work shall focus on the integration of the UK legislation auto categoriser (Google TensorFlow). Suppliers wanting to train a classifier or develop a capability will not be considered.
Nice-to-have skills and experience
  • Knowledge of digital forensics procedures
  • Experience of differing media devices
  • Knowledge of integration of web image metadata matching services

How suppliers will be evaluated

How many suppliers to evaluate
3
Proposal criteria
  • Compliance with Functional Requirements
  • Compliance with Non-Functional Requirements
  • Value for money
Cultural fit criteria
  • Knowledge sharing and participation in skills transfer to the client.
  • Ability to work collaboratively with the client-side team, stakeholders and suppliers.
  • Flexibility in adapting to meet changing priorities and business requirements.
  • Ability to deliver within an Agile framework
Payment approach
Fixed price
Assessment methods
  • Written proposal
  • Presentation
Evaluation weighting

Technical competence

50%

Cultural fit

10%

Price

40%

Questions asked by suppliers

1. Can the Authority please confirm whether any discovery or proof of concept work has been carried out in this area already?
No Discovery or Proof of Concept work has previously been commissioned for the DME.
2. Please could you confirm the deliverable of this project? Is the Home Office looking for a (new) software tool for UKLE with which they can efficiently analyse and categorise vast quantities of media (images, video, text, etc.)
Yes
3. or is this a professional services RFP in which you’re looking for an examiner to attend your London locations (for 24 months) to use and improve the Home Office's existing image-similarity software?
No
4. DME will compare the images against current known images" – Is this to be achieved by hashing all the images found and comparing hash values with those on CAID?
Yes
5. All images will then be run through image-similarity software." – What is the purpose? e.g, if images are classified as legal, why would they be of interest? Else, is it to establish similarity with known illegal images to counter image manipulation designed to alter the hash value and prevent identification?
Yes, and for investigation purposes to aid determination of the offender threat profile.
6. For "image-similarity" – do you want the app to identify and group images and grade similarity?
Yes for image similarity and no for grade similarity, this will be performed by the auto-classifier and other API plugins that the DME provider may have integrations with.
7. .mages will be exported into CAID" – What are the API interface considerations? For example, specific format/compression or in the original format. For large images, will there be a limit in size that then defaults to apply image compression if limit is reached?
The DME will create a case as per the Project Vic OData industry standard json file for ingestion into CAID.
8. Do you have a preference for programming language use? For example, is it a standalone software application you seek or a web-based application?
No preference in language, but the solution must NOT be web-based: images must not go across the internet or intranet for assessment purposes. A secure internal network will be used to transfer the images to CAID.
9. Do you wish the application to sit within the same security environment as CAID e.g ISO27001?
The DME will be deployed within forces, most typically within Digital Forensic Units, which are ISO27001 accredited, and communicate with CAID which sits outside this accreditation.
10. Could you please expand on what you mean by 'Must have a product that is near operational use against the requirements specified in the problem with minimal development and configuration needed' and how this will be assessed?
The supplier must provide a completed product used for digital media cases that has an open API to connect to the auto-categoriser and other services. The Authority will pay for integration and configuration, but not for development of a completely new product.