Matthew Sorell

Conditions for Effective Detection and Identification of Primary Quantisation of Re-Quantized JPEG Images

The choice of Quantization Table in a JPEG image has previously been shown to be an effective discriminator of digital image cameras by manufacturer and model series. When a photograph is recompressed for transmission or storage, however, the image undergoes a secondary stage of quantization. It is possible, however to identify primary quantization artifacts in the image coefficients, provided that certain image and quantization conditions are met. This paper explores the conditions under which primary quantization coefficients can be identified, and hence can be used image source identification. Forensic applications include matching a small range of potential source cameras to an image.

Video provenance by motion vector analysis: A feasibility study

Motion vectors are used in digital video compression for the purpose of reducing file size, primarily by providing an estimate of a frame of interest as a motion-offset version of a reference frame or frames. Despite the name, a motion vector is not in and of itself a trace of object motion, but rather an index to a similar image block. The computational complexity of searching for a good motion vector means that many sub-optimal but computationally feasible approaches have been developed. It is conjectured that proprietary approaches to motion estimation lead to a provenance fingerprint. If the approach can be identified, then the identity of the source firmware or software can be inferred. This has further applications in edit and forgery detection. This paper describes the results of an early feasibility study in which the ability to distinguish between six motion vector algorithms is considered.

Analysis of Sensor Photo Response Non-Uniformity in RAW Images

The focus of this paper is a review of a digital camera identification technique proposed by Lukas et al [1], and a modification of the denoising filter, allowing it to be used for raw sensor data. The approach of using raw sensor data allows analysis of the noise pattern separate from any artefacts introduced by on-board camera processing. We use this extension for investigating the reliability of the technique when using different lenses between the same camera and between cameras of the same manufacturer.

Forensics in Telecommunications, Information and Multimedia

This paper introduces a novel design for handwritten letter recognition by employing a hybrid back-propagation neural network with an enhanced evolutionary algorithm. Feeding the neural network consists of a new approach which is invariant to translation, rotation, and scaling of input letters. Evolutionary algorithm is used for the global search of the search space and the back-propagation algorithm is used for the local search. The results have been computed by implementing this approach for recognizing 26 English capital letters in the handwritings of different people. The computational results show that the neural network reaches very satisfying results with relatively scarce input data and a promising performance improvement in convergence of the hybrid evolutionary back-propagation algorithms is exhibited.

Unexpected Artifacts in a Digital Photograph

This article investigates an unexpected phenomenon observed in a recent digital photograph, in which the logo of a non-sponsoring sports company appears on the jersey of a famous football player in just one of a sequence of images. After eliminating deliberate image tampering as a cause, a hypothetical sequence of circumstances is proposed, concerning the lighting, dominant colours, infrared sensitivity, optical preprocessing, image enhancement and JPEG compression. The hypotheses are tested using a digital SLR camera. The investigation is of interest in a forensic context, firstly as a possible explanation in case such a photograph is observed, and secondly to be able to confirm or refute claims of such artifacts put forward claiming that a hypothetical image is not really what it claims to be.

Embedded importance watermarking for image verification in radiology

Digital medical images used in radiology are quite different to everyday continuous tone images. Radiology images require that all detailed diagnostic information can be extracted, which traditionally constrains digital medical images to be of large size and stored without loss of information. In order to transmit diagnostic images over a narrowband wireless communication link for remote diagnosis, lossy compression schemes must be used. This involves discarding detailed information and compressing the data, making it more susceptible to error. The loss of image detail and incidental degradation occurring during transmission have potential legal accountability issues, especially in the case of the null diagnosis of a tumor. The work proposed here investigates techniques for verifying the voracity of medical images - in particular, detailing the use of embedded watermarking as an objective means to ensure that important parts of the medical image can be verified. We propose a result to show how embedded watermarking can be used to differentiate contextual from detailed information. The type of images that will be used include spiral hairline fractures and small tumors, which contain the essential diagnostic high spatial frequency information.

Signaling Requirements for Smart Dynamic Pricing in Cellular Networks

Smart pricing is a dynamic pricing scheme which varies price in response to the relative scarcity of network resources and has been developed by a team including the authors to be implemented on 3G mobile telecommunications networks. This paper analyses the necessary signaling requirements and proposes a signaling model for smart pricing. In this model, a new network element called dynamic pricing engine is proposed. This element is responsible for setting tariffs depending on the level of congestion in a cell and specifies the quality of service required. Simulations have yielded relationships between the arrival rates for multiple services (such as voice calls and video conferencing) and signaling load. We are also able to estimate how much of that signaling volume results in billable events (so-called cost-recoverable signaling) versus signaling which does not, such as abandoned call attempts (cost non-recoverable signaling) over a range of traffic volumes.

Digital Camera Photographic Provenance

The term provenance is traditionally applied to works of art, referring to documentation which relates to the ownership and public visibility of a particular work, but also includes documentation of production, restoration, thefts, expert opinions on condition and valuations, and any other records which help to assess its integrity. In the realm of criminal evidence, the term chain-of-evidence carries a similar meaning, although the context is usually limited to establishing the source of a particular piece of evidence and its chainof-custody, an even more specific term which refers to the documentation and handling of a piece of evidence once it has been received by investigators. The difficulty with using the chain-of-evidence term for digital media is that the term implies a single version from a particular source, which survives as a single object in custody. Where the chain is AbsTRACT

Topology Reconstruction and Characterisation of Wireless Ad Hoc Networks

Wireless ad hoc networks provide a useful communications infrastructure for the mobile battlefield. In this paper we apply and develop passive radio frequency signal strength monitoring and packet transmission time profiling techniques, to characterise and reconstruct an encrypted wireless networks topology. We show that by using signal strength measurements from three or more wireless probes and by assuming the use of carrier sense multiple access with collision avoidance, for physical layer control, we can produce a representation of a wireless networks logical topology and in some cases reconstruct the physical topology. Smoothed Kalman filtering is used to track the reconstructed topology over time, and in conjunction with a weighted least squares template fitting technique, enables the profiling of the individual network nodes and the characterisation of their transmissions.

Multiple medical image ROI authentication using watermarking

Medical images are now almost all gathered and stored in a digital representation for easy transmission and archiving. High resolution is mandatory for a detailed diagnosis, which requires accurately known location and density information regarding the important features of the image called the regions of interest (ROI). Such features may include non-displaced fractures or small tumors that can often be difficult to identify. A reduction in size by using compression is necessary for efficient transmission over a wireless link where remote diagnosis may be an only option in many cases. Despite rapid advances in lossy compression, most research in the compression of medical imagery specifies that the ROI must be conserved as much as possible or compressed with a lossless or near-lossless algorithm. To ensure diagnostic integrity of these crucial regions after transmission, a multiple watermarking technique has been developed which can be used to verify the integrity of the ROI prior to diagnosis. This has the benefit of assuring that incidental degradation has not affected any of the crucial regions. A strong focus is placed on the robustness of the watermarking technique to JPEG compression as well as the issue image file size and quality tradeoff. The most useful contribution in our work is assurance of ROI image content integrity after image files are subject to incidental degradation in these environments. This is made possible with extraction of DCT signature coefficients from the ROI and embedding multiply in the Region of Backgrounds (ROB).