📸🎙️ Shocking: Robot vacuums caught secretly collecting photos and audio to train AI, raising major privacy concerns.

Ecovacs, a Chinese home robotics company, has been collecting photos, videos, and voice recordings from its Deebot robot vacuums to train its AI models, despite users not being informed of the data collection. The company’s privacy policy allows for the blanket collection of user data for research purposes, including 2D or 3D maps of users’ homes, voice recordings, and photos or videos taken by the device’s camera.

Data Collection and Anonymization

When users opt into Ecovacs’ “Product Improvement Program” through the smartphone app, they are not told what data will be collected, only that it will help improve product functions. The company’s privacy policy states that even deleted data may continue to be held and used. Ecovacs claims to anonymize user information at the machine level, but cybersecurity experts question the effectiveness of this measure.

Cybersecurity Flaws

Ecovacs models have been found to be vulnerable to hacking from over 100 meters away, raising concerns about the security of the collected data. Cybersecurity researcher Dennis Giese reported critical flaws to the company last year, which it has committed to fixing in November. The company’s ability to protect user data is in doubt, leaving it vulnerable to corporate espionage or nation-state attacks.

AI Training and Data Collection

Ecovacs engineers have described the need for large amounts of data to train their AI models, which they have collected from various sources, including user data. The company’s AI models are used to improve the functionality of its robot vacuums. However, the collection of intimate photos and audio recordings raises privacy concerns, as seen in a similar incident with iRobot devices in 2022.

Privacy-Preserving Camera Technology

Researchers at the Australian Centre for Robotics have developed a privacy-preserving camera that scrambles images before digitization, preventing remote access to raw imagery. This technology could prevent similar data breaches in the future. While not yet commercially available, it highlights the need for better data protection in smart home devices.

Conclusion

Ecovacs’ data collection practices raise questions about user privacy and the security of sensitive information. As smart home devices become more prevalent, it is crucial to ensure that companies prioritize data protection and transparency in their product improvement programs.

(Citation: ABC News, “Insecure Deebot robot vacuums collect photos and audio to train AI,” October 4, 2024, by Julian Fell)

Chinese state-backed hackers, known as Salt Typhoon, have reportedly breached US wiretap systems to gain access to broadband networks, including those of Verizon, AT&T, and Lumen Technologies. The hackers may have accessed a vast collection of internet traffic, including court-ordered wiretaps, for months.

The Breach

The Wall Street Journal investigation revealed that Salt Typhoon exploited government systems to access the interception portals required by US law. This allowed them to intercept communications from network users, including those with court-ordered wiretaps for national security purposes. The US has pressured telecoms companies to develop lawful interception infrastructure, but this access can also be exploited by threat actors.

Chinese Denials

China has denied the allegations, claiming the US has fabricated a false narrative to frame China. Salt Typhoon has been linked to other Chinese state-sponsored groups, including Flax Typhoon, Volt Typhoon, and Brass Typhoon, which work together to achieve geo-political goals, such as a potential invasion of Taiwan.

Surveillance Concerns

The existence of wiretap systems means they can be compromised by hackers or tech companies without users’ consent. As technology becomes more integrated into daily life, surveillance capabilities expand. Telecoms firms must build backdoors for law enforcement, potentially turning devices into “spy machines.” This raises concerns about government surveillance, both within the US and against foreigners overseas.

Mitigating Measures

To protect against surveillance, consider using a VPN service to keep your information private. The best VPNs can help mask online activity and prevent monitoring.

Conclusion

The alleged breach highlights the risks of wiretap systems and the need for better security measures. As technology advances, it’s crucial to balance national security with individual privacy.

(Citation: TechRadar Pro, “Chinese hackers allegedly hit US wiretap systems to hit broadband networks,” October 7, 2024, by Ellen Jennings-Trace)

A Dutch driver, Tim Hansen, received a $400 fine in November 2023 for using his phone while driving, but the real reason was that an AI-powered traffic camera system mistakenly identified him scratching his head as phone use. The monocam system, which analyzes images to detect phone usage, misinterpreted Hansen’s innocent gesture due to a lack of training data and nuance.

The Incident

Hansen was surprised to receive the fine, as he hadn’t used his phone that day. Upon reviewing the photo evidence, he saw his hand near his ear but not holding a phone. The AI system and human reviewer failed to recognize the error, leading to a wrongful fine for an offense he didn’t commit.

How Monocam AI Systems Work

These systems analyze images of drivers to detect phone use, trained on datasets of phone usage examples. However, limited or unrepresentative training data can lead to mistakes, as Hansen’s case illustrates. The AI likely flagged his hand position based on patterns learned from other examples, failing to differentiate between scratching his head and holding a phone.

Limitations of AI in Law Enforcement

Hansen’s experience highlights the limitations of relying solely on AI for law enforcement. While AI can minimize human error, it’s not perfect and requires human oversight to prevent false positives. This case raises questions about the risks and responsibilities of AI deployment in public systems and the need for more accurate datasets and better-designed algorithms.

Balancing AI and Human Judgment

The incident demands a balance between AI efficiency and human judgment. AI can process vast data quickly, but struggles with subtle variations in human behavior. A fine balance between the two is necessary to ensure technology serves justice fairly. Hansen has appealed the fine, which may take 26 weeks to resolve.

(Citation: AI Gopubby, “How an AI Traffic Camera Fined a Driver $400 for Just Scratching His Head,” September 27, 2024, by Ritvik Nayak)

Ascension, one of the largest US health systems, has been hit by a ransomware attack that has compromised patient data. The attack, which occurred in July 2024, affected 20 hospitals and 500 clinics, potentially exposing sensitive information such as medical records, financial data, and Social Security numbers.

The Attack

The ransomware, identified as BlackCat, was used to encrypt data on Ascension’s systems. The attackers demanded a ransom in exchange for the decryption key, but the company refused to pay. Ascension has not disclosed the extent of the data breach or the number of patients affected.

Patient Data at Risk

The compromised data includes medical records, billing information, and Social Security numbers. The attack could have exposed sensitive information for millions of patients. Ascension is working with law enforcement and cybersecurity experts to investigate the incident and mitigate the damage.

Response and Mitigation

Ascension has taken steps to contain the attack, including disconnecting affected systems and engaging third-party incident response firms. The company is also offering free credit monitoring and identity theft protection to affected patients. However, the full extent of the breach and the potential impact on patients remain unclear.

Ransomware Threats to Healthcare

Ransomware attacks on healthcare organizations are increasingly common, as they often have large amounts of sensitive data and are seen as lucrative targets. The BlackCat ransomware group has been linked to other high-profile attacks, including on the Irish health service and a US hospital chain.

Conclusion

The Ascension attack highlights the growing threat of ransomware to healthcare organizations. It underscores the need for robust cybersecurity measures and incident response plans to protect patient data. Patients affected by the breach should monitor their accounts and credit reports closely.

(Citation: Perplexity AI, “The Ascension Hospitals Ransomware Attack,”)

Researchers have discovered a new attack method called PixHell that can extract sensitive information from LCD screens by analyzing the noise emitted by the display. The attack works by recording the high-frequency sounds produced by the screen and using machine learning algorithms to decode the audio into text.

How PixHell Works

PixHell exploits the fact that LCD screens produce unique noise patterns when displaying different characters. By recording the audio and using machine learning models to analyze the patterns, attackers can reconstruct the text on the screen. The attack works on a wide range of devices, including laptops, smartphones, and tablets.

Sensitive Information at Risk

PixHell can extract passwords, PINs, and other sensitive data from the screen. The attack is particularly effective on devices with high-resolution displays and high refresh rates, making it a significant threat to modern devices. The researchers demonstrated the attack on a MacBook Pro and an iPhone, successfully extracting passwords and other information.

Mitigations

To protect against PixHell, users can use noise-cancelling headphones or play white noise while typing sensitive information. Manufacturers can also implement noise-reducing technologies in their displays or use additional security measures like password masking.

Conclusion

PixHell highlights the need for better security measures to protect sensitive information on LCD screens. As devices become more sophisticated, attackers are finding new ways to exploit vulnerabilities. Users should be aware of this threat and take steps to safeguard their data.

(Citation: BleepingComputer, “New PixHell Acoustic Attack Leaks Secrets from LCD Screen Noise,” date unknown)

The Ransomware Threat Persists, But Law Enforcement Strikes Back

Ransomware remains a significant threat, but law enforcement has made progress by adjusting its strategies and taking down major groups. Gareth Owenson, CTO and co-founder of Searchlight Cyber, discusses how this shift in tactics has impacted the ransomware landscape.

The LockBit Takedown: A Turning Point?

The year 2024 has seen notable ransomware takedowns, including the disruption of LockBit, one of the most prominent groups. This success was achieved through international collaboration and innovative methods. Instead of solely targeting technical infrastructure, law enforcement focused on damaging the groups’ reputations, recognizing that credibility is valuable in the Dark Web. Operation Cronos, which targeted LockBit, involved seizing servers, freezing cryptocurrency accounts, and publicly exposing the gang’s leader and affiliate members. This strategy undermined LockBit’s reputation and exposed its vulnerabilities.

The Ripple Effect: A Message to the Ransomware Community

The LockBit takedown had a ripple effect, sending a message that no group is immune to law enforcement. Two weeks later, BlackCat, the second-largest ransomware group, claimed to have been disrupted, but later shut down after a large-scale attack. Although more ransomware groups now operate, there has been a 16% decrease in victims since the second half of 2023, indicating a diversification rather than growth in the ransomware ecosystem.

What’s Next for the Ransomware Landscape?

As the landscape evolves, collecting up-to-date intelligence on ransomware groups is crucial for security professionals. While the threat persists, law enforcement’s new tactics have created breathing room by taking out some of the biggest adversaries. However, the ransomware landscape is diversifying, with more groups emerging and affiliates developing their own tooling. This presents new challenges for cybersecurity teams.

Conclusion

The fight against ransomware hasn’t been won, but law enforcement’s evolving tactics have made a significant impact. By targeting reputations and disrupting major groups, they’ve sent a message that the ransomware underworld is not invincible. As the battle continues, staying informed and adapting to the changing landscape will be key.

Reference: Dark Reading