Low earth orbit (LEO) communication satellite constellations such as Starlink provide wider access to better Internet connectivity in remote locations. LEO networks are significantly more complex than their conventional geostationary and medium earth orbit ancestors, and many details are not well documented. Research into LEO networks therefore often uses simulation, particularly with software-based simulators. Software simulators are not well suited to complex distributed networks with concurrency, however. This paper proposes a largely hardware-based simulation that emulates only the satellite link itself. The proposed system can simulate dynamic LEO-like latency and bandwidth profiles.

The Covid-19 pandemic necessitated a rapid transition to online education, forcing most academic institutions to adopt online assessments as a substitute for traditional, in-person examinations. Many of these online assessments were conducted in an unsupervised setting, with an underlying model that largely relied on the trust that students would maintain academic integrity and adhere to the principles of honest scholarship. Unfortunately, this trust-based approach showed its vulnerabilities, as we observed a significant uptick in incidents of cheating and academic dishonesty across many educational institutions including our own. In an attempt to address these issues, this paper provides an in-depth analysis of the specific cheating behaviours we have identified. We detail how the lack of supervision in online settings has led to creative and highly collaborative cheating schemes that we call local and remote “exam parties,” in addition to the use of online platforms like Chegg.com to contract-cheat. The rise in cheating has been so dramatic that it has led to noticeable grade inflation, skewing academic performance metrics and potentially diminishing the value of educational qualifications. To combat this worrisome trend, we introduce an analysis tool specifically designed to detect signs of potential dishonesty in online assessments. The tool consumes student activity logs from popular Learning Management Systems such as Canvas, and digital assessment platforms such as CodeRunner and Inspera, and analyses these logs to flag suspicious behaviours. Combined with rapid-fire answer submissions and other suspicious timing patterns, the reports generated via the tool help us identify collusion and potential contract-cheating. In addition to the tool, the paper discusses preventive measures that institutions can take to minimize the risk of cheating. These include designing assessments to minimize the ability to cheat, enforcing remote proctoring services, and employing individualized assessments. By combining comprehensive analysis, technological solutions, and preventive recommendations, this paper aims to provide educators and institutions with the tools and knowledge they need to uphold academic integrity in this new era of online education.

LZ-End is a variant of the LZ77 compression algorithm which allows random access to the compressed data. In this paper, we use the random-access capability of LZ-End to perform random edits on the compressed data.

The Internet's main transport protocol, TCP, struggles to make full use of the capacity of geostationary (GEO) and medium-earth orbit (MEO) satellite links due to their long latency and bottleneck bandwidth, in particular if these are shared by many parallel TCP connections. Large TCP transfers may back off to effective rates orders of magnitude below the link's available bandwidth. Conventional mitigation measures either violate the Internet's end-to-end principle or contribute to bufferbloat. A hardware-based simulation of coded tunnels in this paper shows that there is another way: Using the erasure-correcting properties of random linear network codes, one can increase both goodput of large transfers and overall link utilisation in both GEO and MEO settings without extra buffer or end-to-end principle violations.

Using the same assessment questions for all students is problematic, especially in large classes. In exams, it incentivizes students to glean answers from neighbours or memorize answers to past questions. In homework, it encourages copying and the use of cheating websites. Individualized assessments remove these incentives and encourage concept-oriented discussions among students rather than mere memorization and exchange of answers. They render copying ineffective, and make posts on cheating websites traceable, effectively mitigating collusion and plagiarism. However, individualized assessments are traditionally expensive to create and grade for all but small classes. Dividni (www.dividni.com), our free framework for creating individualized assessments, reduces this expense and creates a unique assessment for each student in classes of any size. It also enables auto-grading. We use Dividni at the University of Auckland for printable homework, online assessments and examinations, all individualized, for some of our largest classes in computer systems, programming, networking, and cybersecurity, ranging from CS1 to CS3. This demo is a tutorial introduction to using Dividni for creating individualized assessments. We will demonstrate three individualized assessment contexts: printable homework, online assessments, and examinations. The audience will have the opportunity to create their own printable homework and multiple-choice assessments. Bring a laptop with the latest .NET framework or Mono installed. Basic knowledge of HTML and C# will be useful.

The Transmission Control Protocol (TCP) and the User Datagram Protocol (UDP) are the two main traditional transport protocols on the Internet. TCP's congestion control algorithms allow TCP senders to "back off" when they detect packet loss, a feature that UDP does not implement. TCP is thus generally regarded as being "fair" towards other TCP connections, in contrast to UDP. Internet Service Providers (ISPs) with upstream narrowband satellite connectivity often block UDP on their satellite links as it is perceived to be disruptive in the presence of TCP. This paper argues that in practice, many TCP flows on satellite links are too short-lived to back off, and that therefore, these TCP flows behave like UDP. This poses the question as to how much is gained by blocking UDP, a transport protocol that has gained in importance in recent years with the emergence of QUIC and UDP-based peer-to-peer applications.

Academic dishonesty is a widely acknowledged problem in tertiary education, and a range of safeguards and tools exist to both deter and detect cheating. During tests and examinations students are often monitored closely by invigilators, however incidents of misconduct still occur. Multiple-choice questions (MCQs) are particularly susceptible in this regard due to the simplicity with which answers can be communicated. Nonetheless, they are commonly used in large classes due to their convenience. In this work we investigate multiple-variant MCQs, where answer options are randomly selected from predefined sets. This format is less prone to student cheating and overcomes several weaknesses of similar methods. We evaluate this approach in two large computer science courses through student questionnaires and simple item analysis. We find that multiple-variant MCQ exams are perceived as an effective way to reduce cheating behavior, however not all question variants are equally difficult. This introduces concerns around fairness and we discuss approaches to mitigate these issues in the future.

Satellite Internet links to Internet Service Providers (ISPs) in remote locations represent a bottleneck when the link transmission rate is lower than that of the networks connecting to the link at either side. This presents a significant challenge to Internet hosts using the Transmission Control Protocol (TCP) to transfer larger amounts of data: TCP congestion control relies on the detection of packet losses. If these losses occur at the input to the satellite link during bursty packet arrivals, the TCP senders' responses are delayed by at least one round-trip-time (RTT) - by which time the congestion situation at the input to the link may have changed completely. The leads to senders either over- or undersupplying the link for much of the time, an effect that is made worse if a larger number of senders are involved. To mitigate this, operators generally configure an input buffer of a size that typically tries to match the bandwidth-delay product (BDP) of the link. This design rule was also commonly used for Internet routers until it was discovered that it led to "bufferbloat" with standing queues that never drain, and routers are now typically configured with much smaller buffers. However, this does not seem to be the case in the satellite world. Our paper examines the effect of buffer size on queue behaviour and TCP performance using a large hardware-based simulator with a real-world flow size distribution.

LZ77, LZ78 and other compression algorithms use derivates of the original Lempel-Ziv (LZ) parsing algorithm, which computes the LZ production complexity. The focus of this paper is on the original algorithm. Like its derivates, it progressively factors an input string s into a series of steps, each of which consists of a reference to a substring starting in the already parsed part of the s (reproduction) and a single symbol (innovation). In all LZ parsers, the input string generally ends on an incomplete step part-way through a reproduction rather than on an innovation symbol. This LZ parsing tail is usually given scant treatment as it tends to be short compared to the string length |s|, even though its length is bound only by |s|- 1. This paper examines the expected length of the tail for |s| produced by an i.i.d. Bernoulli source, and uses this to derive a new result on the statistics of subsequences of large strings.

This paper considers capacity and delay constraints for coded tunnels across an erasure channel which occurs on shared Internet satellite links. Such links are long latency bottlenecks with a limited memory input queue which drops packets when it overflows. The latency delays ARQ ACK feedback to senders, making it difficult for them to tune their packet transmission rate. This can cause the input queue to oscillate between empty and overflow. Queue oscillation leaves the link underutilised during the empty phases and slows down large packet flows. Channel coding can in principle provide goodput improvement in this scenario by letting senders accelerate to higher packet rates before burst losses occur and by mitigating exponential backoff after losses. However, this is only possible if the codes preserve sufficient spare channel transmission rate for the improved goodput to expand into. We formulate rate and delay constraints that such block codes must meet. Using loss data obtained on a purpose-built simulator network, we show that such coding is feasible in a practical scenario and that partial unit memory (PUM) codes are particularly suitable for this task. In this context, we propose a part-systematic encoding for PUM codes, which performs slightly better than non-systematic encoding.

Conceptionally the User Datagram Protocol (UDP) should be well-suited for real-time applications, e.g., for Voice over IP (VoIP). However, many such applications, e.g., Skype, use the Transmission Control Protocol (TCP) either as a primary protocol or as a backup protocol when UDP is blocked, despite TCP's flow control-related data delays. This paper proposes a technique for the estimation of the application buffer requirements of such TCP-based applications and the amount of data congestion in real-time TCP data streams. We apply this technique to data collected from a global network exchanging synthetic real-time traffic over TCP. Our results show that the buffering requirements vary widely with time and path but can be substantial in many cases.

A locally decodable source code (LDSC) allows the recovery of arbitrary parts of an unencoded message from its encoded version, using only a part of the encoded message as input, a challenge that arises when searching within compressed data sets. Simple source codes such as Huffman codes or Lempel-Ziv compression are not well suited to this task: A decoder starting at an arbitrary point within the compressed sequence generally cannot determine its position with respect to the boundaries between encoded symbols, or requires information found before the starting point in order to be able to decode. In this paper, we propose the use of subsets of self-synchronising variable-length T-codes as source codes and show that local decoding is feasible and practical using subsets of T-codes with bounded synchronisation delay (BSD).

This paper presents flott, a fast, low memory T-transform algorithm which can be used to compute the string complexity measure T-complexity. The algorithm uses approximately one third of the memory of its predecessor while reducing the running time by about 20 percent. The flott implementation has the same worst-case memory requirements as state of the art suffix tree construction algorithms. A suffix tree can be used to efficiently compute the Lempel-Ziv production complexity, which is another measure of string complexity. The C-implementation of flott is available as Open Source software.

The Titchener T-complexity CTof a string has applications in, e.g., randomness testing, event detection and similarity comparison. Like the Lempel-Ziv production complexity, the upper bound of CTis demonstrably not a linear function of the string length. Knowledge of the bound for a given length is however required in order to convert CTinto a measure with linear upper bound such as Titchener's T-information. For this reason, the upper bound of CThas been investigated before by several authors, with various asymptotic solutions proposed. We present a new analytic closed-form asymptotic upper bound for CTbased on the Hurwitz-Lerch zeta function.

This paper proposes an authentication protocol between two parties. The protocol uses a shared secret key, representing a large variable-length code and a lookup table representing a permutation. The authenticating parties exchange randomly generated messages. Both challenge and response messages consist of multiple codewords from the code, and these messages are each truncated to a fixed number of symbols. The actual information to be recovered by the respondent are the codeword lengths used in the challenge message. These lengths are re-encoded to different lengths via the lookup table, and randomly chosen codewords with these re-encoded lengths are used in the response.

The T-augmentation process is used to construct variable-length T-codes. In a previous paper, the author and T.A. Gulliver showed that it may also be used to construct block codes with bounded synchronization delay (BSD codes) as subsets of T-code codewords of a fixed length. However, only the existence of a bound on the synchronization delay was shown, and an actual bound was left as an open problem. The present paper formulates such an upper bound. The paper also discusses whether some of these subsets may represent maximal comma-free codes.

It is well known that any two non-identical codewords from a fixed-length code must belong to different non- periodic equivalence classes under cyclic shift for the code to have a bounded synchronization delay (BSD). Subsets of fixed- length codewords from T-codes, a family of variable-length codes (VLCs) known for strong statistical self-synchronization and a well-explained synchronization model, also follow this rule, albeit with some periodic exceptions. This paper shows how the T-code construction algorithm may be used for the simple construction of maximal block codes with bounded synchronization delay. These self-synchronizing codes are subsets of T-codes and thus resynchronization under decoding with a T-code decoder is both automatic and detectable.

T-decomposition maps a finite string into a series of parameters for a recursive string construction algorithm. Initially developed for the communication of coding trees (M. R. Titchener, June 1996), (U. Guenther, Feb. 2001), T-decomposition has since been studied within the context of information measures. This involves the parsing of potentially very large strings, which in turn requires algorithms with good time complexity. This paper presents a T-decomposition algorithm with O(n log n) time and space complexity

This paper compares the similarity-detection abilities of two string parsing algorithms from the Lempel-Ziv family and the T-decomposition algorithm proposed by Titchener against the Hamming and Levenshtein measures. Our results show that LZ and T-decomposition based measures work in a wider range of contexts. We also argue that T-decomposition based measures represent a good compromise between accuracy and time complexity.