Due to perceptions of efficiency and significant productivity gains, various organisations, including in education, are adopting Large Language Models (LLMs) into their workflows. Educator-facing, learner-facing, and institution-facing LLMs, collectively, Educational Large Language Models (eLLMs), complement and enhance the effectiveness of teaching, learning, and academic operations. However, their integration into an educational setting raises significant cybersecurity concerns. A comprehensive landscape of contemporary attacks on LLMs and their impact on the educational environment is missing. This study presents a generalised taxonomy of fifty attacks on LLMs, which are categorized as attacks targeting either models or their infrastructure. The severity of these attacks is evaluated in the educational sector using the DREAD risk assessment framework. Our risk assessment indicates that token smuggling, adversarial prompts, direct injection, and multi-step jailbreak are critical attacks on eLLMs. The proposed taxonomy, its application in the educational environment, and our risk assessment will help academic and industrial practitioners to build resilient solutions that protect learners and institutions.

Cooperatively creating awareness of the vehicle and its surroundings can improve the safety of the transportation system. Creating such awareness involves frequently sharing the vehicle's location and kinematics information with its surroundings, which can be achieved by broadcasting Cooperative Awareness Messages (CAMs) or Basic Safety Messages (BSMs). The receivers of these messages know the current location and kinematics of the sender and can estimate the possibility of collision. However, continuously receiving CAMs/BSMs allows the receiver to reconstruct the sender's trajectory, in which the full trajectory may reveal information about the users, such as, house and workplace location. Hence, the user's privacy is violated. Prior works focused only on location-based trajectory reconstruction and ignored the other kinematics, such as heading and speed. Ignoring such information could lead to underestimating the adversary who seeks to misuse the communication. This work analyses the privacy loss which arises from additional information in BSMs/CAMs. We propose a trajectory reconstruction model that leverages all kinematics (AKs), including location, heading, and speed. The trajectory reconstruction model is composed of two sub-models, namely, inference and data association models. The first sub-model estimates the probability from the estimated value of a vehicle's kinematics, and the second sub-model performs linking between pseudonyms. We quantify the privacy loss regarding the precision, recall, and F1-score of the ability to identify the correct link between pseudonyms with AK s-based trajectory reconstruction and compare the proposed model with the location-based approach. We also quantify the users' privacy through the uncertainty in the trajectory reconstruction process. We show that, in some scenarios, the AK s-based trajectory reconstruction gains higher precision, recall, F1-score, and certainty in trajectory reconstruction compared to the location-based approach.

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.

Non-Systematic erasure codes provide confidentiality of data and can even provide strong security guarantees with appropriate parameter selection. So far however, they have lacked an elegant and efficient method for keyword search over the codes. While the obvious method of reconstruction before search can be too slow, direct search produces inaccurate results. This has been one of the barriers in the wider adoption of non systematic erasure codes in distributed and secure storage. In this paper we present an elegant solution to this problem by building an index data structure that we call the Search Vector, created from the generator matrix of the erasure code. We show that this method introduces a very small amount of delay in return of a high degree of accuracy in search results. We also analyse the security of the scheme in terms of information leakage and show that the information leaked from the index data structure is very small even when one assumes the worst case scenario of the attacker having access to the Search Vector.

End to end secure data aggregation scheme for wireless sensor networks that are based on public key cryptography generally use elliptic curves. However elliptic curve based protocols require messages to be mapped to elliptic curves before performing any operations and finally reverse mapped to retrieve the message back. No mapping function, however, which is both homomorphic and has an efficient reverse mapping function is currently known. The mapping functions used in many previous protocols require brute forcing to reverse map the message from a point on the elliptic curve. This solution may be feasible on a base station with unlimited energy and processing power but it means that decrypting becomes very inefficient on ordinary sensors. We propose a secure data aggregation algorithm based on bilinear pairing that avoids this problem and makes decrypting data feasible on ordinary sensors.

Sensor clouds are large scale wireless sensor networks (WSNs), built by connecting a number of smaller WSNs together. Each of these smaller individual WSNs may be owned by different owners. Sensor clouds are dynamic in nature, where wireless sensors can be provisioned and de-provisioned for the users on demand. In such a multi-user, multi-owner system, user access control is a significant problem. Previous user access control schemes have been centralized and designed for standalone sensors or smaller networks and do not take large networks into consideration. In large networks, data is generally aggregated in-network during data collection. In this paper, we present a user access control scheme, which unlike other schemes, is distributed and works on aggregated data within a sensor network. Our scheme which is based on attribute based encryption is also able to differentiate between users who require data with the same set of attributes, which would be a necessity in a commercial sensor cloud system. Our scheme gives the flexibility to sensor network owners to control user access of data from their sensors. Finally, we compare our scheme with other closely related schemes in terms of attack resilience and computation and communication overhead to show its effectiveness.

In this paper, we present an efficient and secure code dissemination technique aimed at sensor clouds. Previous code dissemination techniques were geared toward traditional wireless sensor networks. They did not take into account, the dynamic nature of a sensor cloud, where the applications running on the motes may not just be updated but changed completely in successive code disseminations. The technique presented in this paper is based upon the observation that a large amount of code is common between applications in wireless sensor networks. Our technique first discovers the code common across various wireless sensor applications. It then distributes this code in the form of functions a priori into the network. During code dissemination, these common functions are picked up by the sensors from the network. Only a part of the code needs to be transmitted from the base station. This reduces the overall transmitted code and hence the energy consumption. Since, security is important in sensor clouds, we further present a security scheme based on proxy reencryption to provide confidentiality and integrity of the code. We have implemented our scheme using two different proxy reencryption algorithms, on Mica2 and TelosB mote platforms to measure its energy consumption. We have also evaluated our scheme in terms of disseminated code size and bandwidth usage to illustrate its efficiency compared to a popular secure code dissemination technique, Seluge.

With the exponential rise of pervasive computing applications, data privacy has become much more of an important issue than before. When data is aggregated at each hop in a sensor network, it becomes harder to protect its privacy. A number of privacy preserving data aggregation algorithms have recently appeared for wireless sensor networks (WSNs), very few of them however also address the issue of data integrity along with privacy. Data privacy and integrity are two contrasting objectives to achieve in general. In a privacy preserved data aggregation, it becomes easier for an attacker to inject false data hence, we suggest that both privacy and integrity of data should be treated together. In this paper, we present an energy efficient, privacy preserving data aggregation algorithm which also preserves data integrity in WSNs. We analyze the security of the algorithm and provide proofs for confidentiality and integrity. We enhance this algorithm further to localize, to a certain degree, the corrupt aggregator. We provide the results of our implementation of the algorithm on TelosB motes, illustrating that both the computational overhead and the energy consumption are very low. Finally, we compare our algorithm with other schemes having similar objectives demonstrating that our algorithm performs better in terms of band with usage and energy consumption in a WSN environment.

Secure data aggregation in wireless sensor networks has two contrasting objectives, i) Efficiently collecting and aggregating data and ii) Aggregating the data securely. Many schemes do not take into account the possibility of corrupt aggregators and allow the aggregator to decrypt data in hop by hop algorithms. On the other hand using public key cryptography for providing end to end security is not energy efficient. In this paper we present and analyze the performance of the secure hierarchical data aggregation algorithm which uses an efficient public key cryptosystem (elliptic curve cryptography) to achieve end to end security. Unlike many other secure data aggregation algorithms which require separate phases for secure aggregation and integrity verification, the secure hierarchical data aggregation algorithm does not require an additional phase for verification. This saves energy by avoiding additional transmissions and computational overhead on the sensor nodes. We present and implement the secure data aggregation algorithm on Mica2 and TelosB sensor network platforms and measure the execution time and energy consumption of various cryptographic functions. We have also simulated our algorithms to analyze how an end to end scheme increases the network life time. We experimentally analyze our algorithms based on parameters like throughput, end to end delay and resilience to node failures.

Mobile ad hoc networks (MANET) are a subset of Delay Tolerant Networks (DTNs) composed of several mobile devices. These dynamic environments makes conventional security algorithms unreliable, nodes that are far apart may not have access to the other's public key, making secure message exchange difficult. Other security methods rely on requesting the key from a trusted third party, which can be unavailable in DTN. The purpose of this paper is to introduce two message security algorithms capable of delivering messages securely against either eavesdropping or manipulation. The first algorithm, Chaining, uses multiple midpoints to re-encrypt the message for the destination node. The second, Fragmenting, separates the message key into pieces that are both routed and secured independently from each other. Both techniques have improved security in hostile environments. This improvement has a performance trade-off, however, reducing the delivery ratio and increasing the delivery time.

Data aggregation is a technique used to conserve battery power in wireless sensor networks (WSN). When securing such a network, it is important that we minimize the number of computationally expensive security operations without compromising on the security. This paper deals with the test-bed implementation of our end to end secure data aggregation algorithm. Unlike previous algorithms which required separate phases for secure aggregation and integrity verification, ours does not require an additional phase for verification. This saves energy by avoiding additional transmissions and computation overhead on the sensor nodes. We have implemented our secure data aggregation algorithms on mica2 motes.

Data aggregation is a technique used to conserve battery power in wireless sensor networks (WSN). While providing security in such a scenario it is also important that we minimize the number of security operations as they are computationally expensive, without compromising on the security In this paper we evaluate the performance of such an end to end security algorithm. We provide our results from the implementation of the algorithm on mica2 motes and conclude how it is better than traditional hop by hop security.